Subscribe now

A frac bid is one of the more precise documents in oilfield procurement. It quotes you a rate to the dollar: so many dollars per stage, so many per pound of sand, so many per lateral foot, some of it carried out to a third decimal place, as if the third decimal were the thing that would win or lose you the well. The precision is genuine, and it is aimed at the wrong number. The figure that actually sets what a well costs does not appear on the bid to any number of decimals, because it does not appear on the bid at all.

A frac spread is rented by the day. Call it two hundred thousand dollars of it, which is roughly where the public arithmetic lands once you divide a listed pressure pumper’s revenue by its active fleets and its operating days. The real figure runs from about $187,000 at the lean end to nearly $300,000 for the highest-cost fleets, and $200,000 is a fair, round, conservative place to stand. The spread shows up, it pumps 18 to 22 hours a day, and it stays parked on your pad until the well is done. So to a first approximation, the cost of the well is the day-rate times the number of days. That approximation is also the whole game.

Everyone at the table negotiates the day-rate. The number of days barely comes up, which is odd, because the number of days is what actually sets the bill. It is not even a mysterious quantity. It is the length of the lateral divided by how fast the crew pumps it, which means the entire cost of the job rests on a single variable, completion speed, and completion speed is the one number that never makes it onto the bid you are comparing.

It appears nowhere because, until recently, nobody had it. You can put an exact price on a stage, because a stage is a thing you can count on an invoice. Nobody put a price on a day-per-well, because that would mean holding a stopwatch on every crew, on every pad, across years, and then asking the question the invoice never makes you ask: do two spreads with the same logo on the door pump at the same speed?

They do not. They are not close. The gap is not small enough to round away, and it is exactly the size that quietly decides which bid was the cheap one, long after you have signed the other.

This is a piece about that gap. How big it is, why your bid sheet cannot see it, and what it does to the price of a well once you can. For this example, we will use our data set for the DJ Basin’s frac fleets, one fleet at a time, and ran the bids back through it. The cheapest headline number, it turns out, is usually not the cheapest well. You can prove it from the iron up.

It Isn’t the Logo, and It Isn’t the Design

Start with the comparison everyone already makes, the one between the two companies. In the DJ, the two providers we can attribute pump at visibly different speeds: one runs a basin-wide median of 2,362 lateral feet a day, the other 3,198. Call it a 35% gap. That is a real difference, and if it were the whole story, leveling bids would just mean adjusting for it. Dock the slower logo 35% and move on.

That is the whole story only if the logo is the whole story. Look inside either one and the 35% turns out to be the small part.

The slower provider’s own fleets run from 1,468 feet a day at the bottom to 2,815 at the top, a factor of 1.9. The faster provider’s fleets span 2,470 to 4,381, a factor of 1.8. So the gap between the two companies is 35%, and the gap inside a single company is nearly double that. Whatever you think you are buying when you pick a logo, a completion speed is not it.

Here is where someone in the room says the sensible thing, which is that not every job is the same. A fleet pumping a heavy, sand-loaded design will be slower than one pumping a light one, and that has nothing to do with the crew. This is correct, and it is the first thing the data confirms. Sort the basin’s jobs by proppant intensity and the speeds fall in a clean staircase: under 1,000 pounds per foot, the median fleet pumps 3,984 feet a day; at 1,800 pounds and up, 2,280. Heavier designs run about 43% slower, every time, for everyone. Design is a real variable, not a rounding error.

Which is exactly why you cannot compare raw feet per day blind, and exactly why the comparison that matters holds the design fixed. So hold it fixed. Match the fleets on proppant loading, water, and lateral length, and read the crew on the residual. When you do, the gap does not go away.

Look back at the slower provider’s fleets. A-1 and A-3 pump almost identical designs, about 1,100 pounds of proppant per foot, matched on water and on lateral length. A-1 runs 1,468 feet a day. A-3 runs 2,125. That is 45% more well per day, at the same design, under the same logo. Reach across to the faster provider and the matched comparison gets worse: A-3’s 2,125 against B-5’s 3,878, at the same loading, is an 82% gap. And design does not even fix the ordering. B-1 pumps the heaviest program in the basin, 2,084 pounds per foot, and still outruns several of the slower provider’s lighter fleets. If sand loading set the speed, that could not happen.

So the answer to “not every job is the same” is yes, and we account for it, and the gap survives anyway. The residual is the crew, the iron, the pad logistics, the thousand small decisions that make one spread fast and another slow on identical rock. That residual is the thing your bid sheet is built to ignore.

Operators know this, even when procurement does not. Occidental spent late 2023 deliberately cutting its Niobrara proppant loading from roughly 1,500 pounds per foot down to 800, and its Codell designs from 1,100 to 550, to improve capital efficiency and ease the mechanical strain that slows a job down. Civitas flattened its loading and cut fluid 23% to hold its execution steady. Both of them changed the days.

And that is where the money actually sits. Take that same $200,000 day-rate and run it across the basin’s operating band on a 10,167-foot well: the slowest fleet turns it into about $1.39 million of pumping cost, and the fastest turns the identical rate into $464,000.

That is a $921,000 swing on one well, and none of it lives in the day-rate everyone spent the RFP arguing about.

Put thirty of those wells in a program and the day-rate you negotiated so hard stops being the number that decides what the campaign costs. What decides it is the crew that pumps them, and the crew is the one input a bid sheet gives you no way to see.

What Happens When You Price the Days

The standard way to compare two frac bids is to put them in the same units. Dollars per stage, dollars per pound of sand, dollars per lateral foot: pick a denominator, divide, and the smaller number wins. Everyone is taught to do it and it is a reasonable thing to do, and it also quietly prices the rate while ignoring the days, which are the whole ballgame. A crew that finishes in three days and a crew that finishes in five can quote the identical dollar per stage. The bid sheet calls them a tie. The pad does not.

If that sounds like a technicality, the federal government disagrees, and has for decades. Federal procurement rules, the ones written so a contracting officer cannot be fooled by a low headline number, require that competing prices be adjusted for materially differing terms before anyone compares them. Completion speed is a materially differing term. It is, in fact, the term that moves the bill the most. The unit-rate method leaves it out for a dull reason: until recently there was no number to put in that column, so nobody built the column.

We have it now, at depth. Across the Lower 48 we benchmark six attributable providers on hundreds to thousands of jobs each, deep enough that the percentiles actually mean something. The speed distributions are wide, and they stay wide inside a single provider: the spread from a provider’s 25th-percentile job to its 75th runs from under 60% to nearly 120%. A single median hides most of that, and the part it hides is exactly the part you are signing for.

So you stop comparing bids and start pricing them. Take each bid, run it through the should-cost model at that bidder’s actual completion speed, and turn a headline dollar-per-stage into the only thing you actually care about, dollars per well. Do that to a real eight-bidder RFP and the ranking does not survive the trip.

Subscribe now

The cheapest headline bid, the one that would have won, lands second once its speed is priced in. The third-cheapest falls to seventh. The most expensive bid on the sheet, the one procurement would have thrown out first, turns out to be the cheapest well of the eight. The headline ranking and the ranking you actually pay are two different lists, and the one on your desk is the wrong one. This is a point-in-time run on one real Rockies RFP, so read the order, not the third decimal. The order is the point, and it is the same order the live basin data draws: the providers that pump fast stay fast, and the cheap-looking slow ones stay expensive.

The DJ version is cleaner, because there the gap is pure. Take our two providers, quote them the exact same day-rate, and change nothing but the speed each one actually pumps. Provider A delivers the well for $860,881. Provider B delivers it for $635,835.

That is a quarter of a million dollars off the same well, 26% cheaper at an identical day-rate, and every dollar of it was bought with days. Provider B does exactly one thing Provider A does not: it pumps about 35% faster. That is the whole difference. A dollar-per-stage bid has nowhere to record a thing like that, so it doesn’t, and you find out on the pad.

A Cheap Bid Is a Promise Somebody Has to Keep

So far the question has been which bid is actually the cheapest. There is a second question underneath it, and it is the one that decides whether the cheap bid is even real: can the bidder do the job for the price it named, and keep doing it for the length of your program?

Break a normalized bid into what it is actually made of and most of it is two things, pumping and fuel, with not a lot else. Fuel is not a rounding error and it is not a constant. A legacy diesel spread burns something like $7,565 of energy per pumping hour; a dual-fuel fleet runs nearer $3,400; a fully electric one nearer $3,000. The drive-train under the hood moves the whole cost build-up before the crew pumps a single stage, which is why the model prices each fleet’s fuel off the iron it actually runs instead of a basin average.

Underneath the build-up sits a floor. A service provider needs roughly 11% over its true cost to keep crews paid, iron maintained, and the next job staffed. Bid below that for long and something gives, usually in the middle of your program, which is the worst possible time for it to give. Here is the caveat, and it matters enough that hiding it would make the model useless: in this RFP, every bid flags below the sustainable band. That is not a verdict on eight reckless bidders. It is a scope mismatch, a full-fleet cost floor measured against a pumping-only line in the RFP, and the honest move is to say so rather than dress it up. The floor figure itself stays ours.

What survives the caveat is the relative read. The cheapest bids sit closest to the line, and a bid well under sustainable economics is not a discount you get to keep. It is a promise about crews and maintenance, and there are two very different stories that produce the same promise on paper. In one, the bidder runs newer, cheaper iron, genuinely pumps for less, and is handing you the savings to win the work. In the other, the bidder needs the backlog, named a number it cannot actually hold, and is quietly counting on making it back once its spread is the only one left standing on your pad. Both stories show up as the same PDF. From the bid alone I cannot tell you which one you are reading, and neither can your operating team. What tells you is whether the number clears the cost of doing the work. The timing is what makes that worth checking right now: premium fleets sold out through 2026, spot pricing moved up 10% to 30% this spring, and a number that pencils today on a thin margin is the first one to get repriced, or quietly under-crewed, the moment it can be.

A Discount Can’t Buy Back a Day

When you learn one provider is slower, the reflex is to go ask it for a discount. Reasonable. So price the reflex. Fix the faster provider at its median speed, and ask how large a day-rate discount the slower one would need to match it on cost per well.

There is no such discount. At full rate the slower provider would need to pump 3,198 feet a day to match. Cut its rate 10% and it needs 2,878. Cut it a full 20%, which no one is going to give you, and it still needs 2,558. It actually runs 2,362. The line never reaches the bar. A rate concession cannot buy back a speed gap, because speed compounds against every day on the pad and a discount only touches the rate on each of them. Run it against the distribution, the 25th to 75th percentile, not a single point, so you are negotiating the range of outcomes instead of one optimistic guess.

Which is the real move once you can see the days: the negotiation slides off the rate and onto the terms that govern how many days you are buying. The sharpest of them is an out-clause that lets you stop paying full rate while the clock runs slow, which is the same arithmetic that explains why no operator will risk a wellbore to chase a nominal $3,500 a day when the spread on location is already burning two hundred thousand of it. Every lever like it is a clause about time, and time is the one thing the day-rate, argued on its own, does almost nothing to control.

What to Demand Before You Sign

The Q3 RFP is where this gets decided, so here is the short version of what completions and procurement should walk in holding. Score on performance-adjusted cost per well, not headline dollars per stage. Normalize for design before you compare crews, so the heavy program and the slow crew stop looking like the same thing. Verify completion history before you award, and put the thin-history bidders on a pilot pad before you hand them a program. Write the speed floor and the NPT out-clause into the contract, not into the kickoff meeting. Run the sustainable-margin check, so the winning number is one the provider can actually hold for a year.

The Part We Built

None of this works without the measurement, which is the thing we spent the last few years building. The first piece is a fleet-level frac performance feed: roughly 32,000 attributed pad-jobs across about a dozen basins, completion speed at the pad level, refreshed weekly off public data.

The second is a bottom-up should-cost model anchored to the financials of the one public pure-play frac company, built up from pumping, fuel by drive-train, and delivered sand priced to the basin rather than a national average. Put them together and the bid sheet stops being a list of rates and becomes a list of wells, each priced at the speed its bidder actually pumps.

Which means we can do for your RFP what we just did for the one in this piece. Hand us your bid set and your well design, and the model turns each bidder’s rate into a cost per well at the speed that bidder actually pumps, adjusted for your basin’s sand and fluid instead of a national average, and hands you the performance-adjusted re-rank before you sign anything, while the order can still change the award. It is not a one-off memo. It is the same live feed and model this piece runs on, pointed at your bids. If you are running a frac RFP this quarter, the most useful thirty minutes you will spend on it are the ones you spend looking at your own bids re-ranked.

You are not buying stages. You are renting days, at a rate you negotiate and a speed you do not. The bid quotes the rate to the dollar and says nothing about the speed, and the speed is the only variable that decides what the well costs. That was tolerable while the days had no number and everyone was equally blind to them. They have a number now, which means the next RFP you run is the first one where not looking is a choice.

Until next time. Stay hedged.

Share

Subscribe now

A compression vendor makes money when horsepower is set and earning. A drilling contractor makes money when rigs are turning, a frac company when crews are pumping, a tubular distributor when pipe is leaving the yard. Nobody would pay for that observation, and I’m not asking you to. But the entire commercial posture of the service sector hangs on the back half of those sentences: earning, turning, pumping, for somebody. The iron does not care who the somebody is. You’ll do. So will your neighbor.

Most operators prepare for a renewal as if the vendor’s decision is about them, so let’s run the standard preparation and watch it not matter.

Suppose you run 30 units in the Delaware with a compression renewal coming up in Q4, and you’ve done the homework the textbooks assign. You know your BATNA (Best Alternative to Negotiated Agreement) cold. You know what insourcing would cost (roughly a seven-year payout on new iron, which is why it’s a calculation you run and not a plan you execute). You know which competing vendors have idle units in your size class, and how long you could limp along on rentals if talks went badly. You walk in knowing exactly what you’ll do if they say no.

The regional VP across the table has a different spreadsheet open, and you’re not on it. It lists every other operator running his equipment in the basin: who’s adding rigs, who just closed a merger, who got upgraded to investment grade last month, who’s unhedged into a soft strip, whose board locked a multi-year program and said so on the call. If he were being candid, which he is paid specifically not to be, his side of the negotiation would go something like: “Hi. I like you. Your checks clear and your ops team doesn’t yell at my mechanics. But if you walk, this spreadsheet says your horsepower is set at somebody else’s tank battery inside a quarter, at the same rate, possibly with a mobilization fee thrown in for my trouble. So, the rate is the rate.” Notice that nothing in that speech is about you. Your homework never comes up.

The arithmetic underneath every deal we work is one line: your BATNA minus the vendor’s BATNA equals the zone of possible agreement. Operators are, in my experience, genuinely good at the first term and almost never analyze the second. And the second term has two components that get conflated. One is the vendor’s own economics: debt pressure, idle inventory carrying cost, contract expiration cliffs, spot-versus-legacy tariff spread, fund lifecycle if he’s PE-backed. We score that on an 11-dimension Leverage Matrix, per basin rather than per vendor, because the same company that shrugs you off in Midland, where it keeps 65% of its fleet, will bend in the Powder River, where it keeps less than 5%. That component you can read out of a 10-K if you know where the bodies are buried.

The other component is the vendor’s alternatives, and the part that took me embarrassingly long to internalize is that it isn’t a fact about the vendor at all. It’s a fact about your neighbors. The thing that lets him quote “the rate is the rate” with a straight face is the aggregate durability of everyone else’s demand. For years this was the fuzziest number in our whole system. We could read a vendor’s covenants down to the footnotes and still be guessing about the order book standing behind him.

Game theory has had a name for this problem since before horizontal drilling existed. Lloyd Shapley eventually got a Nobel for asking what each member of a group is actually worth when the group creates value together, and the classroom example is gloves. You own a left glove. Two strangers each own a right glove. A pair sells for $100; singles sell for nothing. Average over every order the three of you could arrive and credit each arrival with the value they add at that moment, and the split comes out $66.67 for the left glove, $16.67 for each right. (I would have guessed something like 50/25/25 before running the orderings myself, which is rather the point. Value doesn’t split by contribution to the pair. It splits by scarcity within the group.) Neither right glove got worse. There were just two of them.

A vendor’s book of business is the coalition. He brings the iron, you and every operator around you bring demand, and the value on the table is the utilization the book supports, which in a levered rental business is a solvency metric wearing a performance metric’s clothes. Your rate, your allocation priority, your place in the queue when a 2,500-horsepower package frees up: all of it is a Shapley split being computed slowly, by feel, by regional sales VPs who have never heard of Lloyd Shapley but can tell you to the unit how many right gloves are in their basin. Diamondback’s rig count is in your negotiation whether you invited it or not.

So the second term of the BATNA equation is measurable after all. It just isn’t measurable by studying the vendor. You have to study everyone else’s demand: nominal, relative to yours, scored for durability, across every basin the vendor could redeploy into. That’s what the Activity Resilience Index does. And because vendors already run this ranking informally (every allocation decision of the 2022-through-2024 scarcity was one, made by feel, and mostly made correctly), the same index read from the other chair is a targeting model: which customers to bank on, which to paper up with term before they wobble. This quarter we read every Q1 2026 earnings call in a 60-plus company universe and scored them all.

An Index That Only Moves When Someone Talks

The ARI answers one question per operator: how likely is this company to keep buying oilfield goods and services at or above current levels for the next 24 months? It’s deliberately service-agnostic, because a rig drop cancels the drilling contract, the frac crew, the sand, the water, and eventually the compression expansion all at once. The unit of analysis is the drilling program, not any line item hanging off it.

Seven dimensions, each scored 1 to 10, weighted by how directly they control that answer. Capital commitment durability and financial capacity carry the most weight, because activity dies in exactly two ways, by decision or by force, and those are the two gauges. Inventory depth and hedging come next: inventory tells you whether the program can physically continue, hedging tells you whether the drilling decision is coupled to the strip or insulated from it. Basin economics, operational momentum, and management signaling round out the seven. (The exact weights are ours.) Signaling earns a slot in a quantitative model because tone leads capital by one to two quarters; a CEO who starts saying “flexibility” in February drops the rig in June. The composite bands: 8.0 and above is fortress demand, the kind a vendor’s board will underwrite a fleet expansion against. 6.0 to 7.9 is durable. 4.0 to 5.9 is at risk. Below that, nobody should be underwriting anything against you.

Two mechanical details do most of the work. The index only rescores a company when it files a new earnings call, investor day, or special call; no new information, no new score. That’s the correct cadence for a measure built on management’s own commitments, and it makes the quarter-over-quarter delta the tradeable part, because a falling score is a procurement window with a date on it. And company scores aggregate to a basin score weighted by rig share, and the basin score is the number that prices your deal. Rising basin ARI: the coalition around you is strengthening and your marginal contribution is shrinking. Falling: your volume is becoming the difference between the vendor’s fleet working and the vendor’s fleet parking.

This is, bluntly, an information asymmetry trade. The vendor knows his book; he lives in it. The point of scoring the public tape is that his read of the basin stops being privileged. Forty-six companies triggered rescores this season. Here’s what came out.

The Quarter Everyone Got Stronger

The Permian sits at 8.92, up 0.35 on the quarter, with four companies improving, seven stable, and zero deteriorating. Breakevens of $25 to $35 per barrel and consolidator balance sheets have produced a customer base that vendors of every stripe can underwrite expansion against, and they will. If you’re negotiating anything in the Delaware right now, compression, rigs, sand, your counterparty’s alternatives have never scored higher. (Is that the strongest vendor position in the modern history of the basin? I don’t know, man, the tape only goes back so far. It’s the strongest since anyone’s been scoring it.)

Appalachia is the quarter’s biggest gainer at 8.82, up 0.42, and the mechanism is a demand regime change rather than a price call. CNX and Range carry hedge books above 60%, and the basin is rotating from seasonal price speculation toward baseload contracts feeding data centers and utility-scale power. EQT curtailing 10 to 15 Bcf in Q2 to protect pricing reads, in this frame, as discipline rather than distress. The gas is sold before it’s drilled, which makes the drilling very hard to talk yourself out of.

The Montney and broader WCSB print 8.84, up 0.12, with the deepest stable cohort in the study, fourteen companies. The story is takeaway: Trans Mountain’s 590,000 barrels a day of new egress and LNG Canada’s 2.0 Bcf/d of processing turned a landlocked resource into an export business, and reserve life north of 15 years did the rest. The Eagle Ford (8.58, down 0.10) and the deepwater Gulf (8.50, flat) are stable for opposite reasons: deepwater capital is committed years ahead and doesn’t react to the strip, while the Eagle Ford is a cash engine that diversified owners like ConocoPhillips and Murphy harvest to fund growth elsewhere. Steady, but nobody’s adding.

Then there’s the Haynesville, at 7.58 the weakest major play in the table, with capital durability at 7.00 and financial capacity at 6.00, the two lowest dimension scores anywhere in the study. Pure-play gas operators there are absorbing severe margin compression, and Comstock’s leverage and capital deficit mean the plan flexes with the basis; flexes, in the bad case, into rig drops. It’s the one basin in the table where the scarce glove is on the operator’s hand.

The mid-quarter runs also caught three basins outside the quarter-end table deteriorating: the Bakken at 5.85 (down 0.60, on structural productivity decline), the DJ at 5.35 (down 0.55), and the PRB at 4.75 (down 0.65). Coverage there is thinner and the reads are noisier, but the direction is consistent, and every vendor selling into those basins can feel it without a model.

The Improvers Didn’t Add a Single Rig (Q1)

Seven companies moved more than a full point this quarter, six up and one down, and you could tell the up story as six unrelated pieces of good news:

1. Devon closed the Coterra merger and came out at 0.90x pro forma net leverage with expanded buyback capacity. Up 1.25 points, 7.73 to 8.98.

2. Antero integrated the HG acquisition and collected its first investment-grade rating. Up 1.18.

3. Tamarack Valley sold Charlie Lake and turned into a debt-free Clearwater pure-play. Up 1.15.

4. SM Energy retired its most expensive debt with a $950 million South Texas divestiture. Up 1.15.

5. Ovintiv integrated NuVista with net leverage under 0.80x. Up 1.08.

6. Permian Resources landed a triple investment-grade rating with well costs down to $685 per lateral foot. Up 1.05.

Now count the rigs in that list. There aren’t any.

So one available reading of the quarter is that nothing happened: same rigs, same volumes, some paper got shuffled, fine. The more informative reading is that the thing vendors actually price moved a lot. The quantity of demand barely budged; what moved is how much of it a vendor’s board, or a vendor’s lender, will believe. An investment-grade consolidator levered under 1.0x is demand you can underwrite a fleet expansion against; the same rigs run by a stretched single-basin operator are a hope with a decline curve. Q1 2026 converted a large block of North American drilling from demand vendors wished for into demand they can bank, and in the gloves game this is the quiet move where the right gloves organize. Your program is exactly what it was in March. Your marginal contribution fell anyway, because the demand around you got more certain.

Consolidation deserves its own line, because it transfers negotiating power twice. When Devon absorbs Coterra, the basin’s aggregate demand gets more durable, which strengthens every vendor’s hand. And the combined company becomes a much larger share of any single vendor’s book, which strengthens Devon’s hand. If you’re a 20-unit independent in the Delaware, both transfers came out of your side of the table, at closings you were not invited to. I mean, congratulations to everyone involved.

The one decliner is geology voting. ARC slipped 0.10 after upper Montney wells at Attachie underperformed and management paused the ramp to re-evaluate, which makes the Attachie corridor the only soft demand in an otherwise firm WCSB. Small number, real window, and somebody in the wet-gas corridor is going to get a discount out of it.

Where Your Volume Still Matters

For operating teams, the playbook falls straight out of the table, and it applies to the whole service stack, not just our corner of it.

In the Permian, Appalachia, and the WCSB, your marginal contribution is shrinking, so stop negotiating like it isn’t. Asking for rate cuts in an 8.9 basin is asking the landlord to mark down the building; in compression especially, the rate is the vendor’s book value, and he’ll hold it because his auditors and his covenants require him to. Lock allocation early, take the term, and extract value where it doesn’t reprice the asset: mobilization credits, guaranteed response times, spare-part consignment, performance clauses with teeth. The same logic runs for rig lines and frac calendars, where the concession currency is crew quality and schedule priority rather than steel. Should the insourcing math still be in your prep? Yes, obviously; the calculation is the BATNA even when the execution isn’t. Will it move the rate in an 8.9 basin? It will not.

In the Haynesville, and in the Bakken, DJ, and PRB if the mid-quarter reads hold, you are the utilization. Act like it: spot pricing, structural discounts, extended payment terms, consignment inventory. The mid-quarter runs caught Appalachian operators playing the same hand from the other side, timing completions around weak price windows and taking cost concessions across nearly the whole vendor list. Leverage is perishable, and in a falling basin every quarter you wait to renegotiate costs the vendor more than it costs you. Both sides’ models know it.

And everywhere: watch dimension four. Hedging is the soft spot of the entire study. The Permian averages 8.36 on it, the WCSB 7.84, the Eagle Ford just 7.10. Magnolia scores 6.50 because it chooses to run unhedged; Murphy’s 5.00 is the lowest hedge score in the table. Tourmaline has already built the option, a $200 million deferral that lets it pause completions and push out up to 60% of well costs if AECO stays weak. Fortress scores are conditional on the strip in a way the composite politely averages away.

For vendors and their investors, the same table read upside down is a targeting model, and this holds whether you rent horsepower, rigs, or crews. Rank your book by ARI. The 8.5-and-above names get allocation priority and your best iron, because their demand is what you take to your board. The fragile tail gets term-length protection, not handshakes. And weight dimension seven more than feels natural, because the customer who starts saying “optionality” on calls is telling you, two quarters early, what the renewal conversation is going to sound like.

Casing Doesn’t Do Investor Relations

A fair objection to all of the above: ARI is built from the tape, and the tape is talk. Management signaling leads activity by a quarter or two, which is exactly why we score it, but managements also spin, and a score assembled from earnings calls inherits whatever optimism survived the general counsel’s markup. You know this. You’ve helped draft the language. If you wanted to know what a basin is actually doing, as opposed to what it’s telling analysts it plans to do, you wouldn’t listen. You’d watch the steel.

The steel is the other half of the system, and the half this article isn’t giving away. Kalibr’s data subscription tracks the physical layer of the same demand ARI scores financially, and the coverage runs to roughly 80% of a drilling program’s CAPEX and 80% of its OPEX. (Pareto would approve.) Three streams do most of the work.

Casing demand, by string. Rig count was a fine demand proxy when a rig meant a predictable number of wells, and that world is gone: laterals stretched past three miles, cycle times fall every quarter, and one modern rig does the work 2014 assigned to three. Casing feet by string is the unit of development that efficiency can’t deflate, because every lateral foot gets cased whether the rig count rose or fell. Track it by area and you’re watching production centers form before anyone writes a press release about them.

Frac performance, at the vendor fleet level. Fleet count has the same disease rig count has (simulfrac and e-fleets mean fewer fleets complete more wells), so throughput per fleet, per vendor, is the number that tells you how much production is actually coming online. Read the other way, it tells you which pumper’s calendar is genuinely full and which one is quoting confidence, which is the vendor-BATNA question again, answered in stages instead of statements.

Compression mapping. Once production exists it has to move, and horsepower is where volumes declare themselves. Map the compression and you’re mapping the flow: which gathering corridors are filling, which are saturating, where the next bottleneck forms and who’ll be paying for it.

One more thing about those streams, mentioned quietly because it’s the practical part: the steel comes with names on it. Casing tracks to who sold it, frac throughput to whose fleet pumped it, horsepower to whose nameplate is on the package. So the physical layer is more than a map of the basin. It’s a map of who does business with whom, in every one of those categories. You’ll remember the regional VP’s spreadsheet from the top of this piece, the one listing every other operator in his book, the one you’re not on. The pairing data is, functionally, that spreadsheet, except it also covers the books of every vendor he competes with. What you’d do with that at a renewal I’ll leave to you.

Each of these is a major cost center first, and the line items justify the read on their own. Stack the three and you get a map with a time axis: casing shows where production is heading, frac shows how much is coming, compression shows where it’ll flow. ARI tells you what a basin says it will do; the steel tells you what it’s already doing; the interesting negotiations live in the gap between the two. I genuinely don’t know how many of your counterparties are watching both sides of that gap. My guess is one or two of the large ones, imperfectly. If you’d rather be on the better-informed end of that particular asymmetry for once, the score in this article is the free half of the pairing, and the reply button covers the other half.

The Crowded Top of the Table

Anyway, the delivery mechanics. Kalibr runs the ARI on an earnings-call gate: a company files a transcript, the score updates, the basin aggregates recompute, and anything moving a full point gets flagged the same week. Subscribers get the rescores as they happen, the full dimension-level scorecard each quarter translated into basin-by-basin negotiating posture, and the physical-layer read alongside it. This quarter’s full table, all 46 companies, all seven dimensions, ships with this issue.

What nags at me is the top of the table. Forty-two of the 46 companies sit in the fortress band, 8.0 or above. EOG tops the list at 9.46, and even the bottom of the table would have looked enviable three years ago. Is that a golden age of customer quality? If you sit on a vendor’s board, sure, and vendor boards are acting like it. Is it also exactly what a table built partly on half-hedged books would print at this point in the strip? Also yes. The crowd’s weakest dimension is hedging, which means a 15% slide in the strip wouldn’t trim this table so much as re-sort it, and next quarter’s movers list would fill with names heading the other direction. Whether today’s scores are a durable regime or a high-water mark isn’t something the current filings can tell you. The next earnings-call gate opens in October. We’ll run the tape and find out.

Subscribe now

Share

Nominally Hedged is published by Kalibr Partners. Reply, or message us directly, and we will point you to the report built for your side of the table: the quarterly ARI scorecard for operators, the customer-durability targeting read for compression and oilfield service vendors.

This newsletter is research and commentary for informational purposes only. It is not investment advice and not a solicitation to buy or sell any security. The analysis draws on public filings, earnings call transcripts, and investor presentations alongside Kalibr’s proprietary scoring framework; figures are believed reliable but not guaranteed, and reflect modeling assumptions described in the piece. Company names and marks belong to their respective owners.

Subscribe now

The most useful thing Annie Duke took from twenty years at a poker table was not a tell, or a bluff, or a way to count the cards that could still save a hand. It was a complaint about the English language. We do not have a clean word, she noticed, for a good decision that happens to turn out badly. When the cards come in wrong we call it a mistake, and when they come in right we call the player across the table a genius, and most of the time we are wrong on both counts, because we are grading the dealer’s work and crediting it to the player. She named the error “resulting”: judging the quality of a decision by the quality of its outcome, two things that are correlated, on average, eventually, and almost never in the moment you actually have to decide.

The cure she prescribed was to stop thinking in answers and start thinking in bets. A decision is a bet on a future you cannot see. When you choose, you are not declaring what is true; you are pricing what is likely, putting capital behind a probability, and accepting that a good price can still lose. This is not a metaphor energy people need translated. An oil and gas company is a machine for making bets under uncertainty, and everyone who works inside one knows it, whatever the investor deck calls it. You drill a well against a type curve you will not confirm for eighteen months. You hedge a strip against a price you do not control. You sign a ten-year midstream dedication against a basin differential nobody can forecast past next winter. The drill bit is a wager. The hedge book is a wager. The acreage is a wager. The only unusual thing about the industry is how rarely it says the word out loud.

The reason to say it out loud now is that the rest of the world has started pricing beliefs the same way, on purpose, and charging admission. Prediction markets have gone from a novelty to a quote you check, and the thing they do that a press release cannot is strip a decision down to its implied probability. A contract trading at sixty cents is not an opinion. It is a number that says the crowd thinks this happens sixty percent of the time, and you are welcome to disagree, except now your disagreement has a price and a counterparty. Once you have spent any time reading the world that way you cannot stop, which is the small tragedy of learning it. Every commitment starts to look like a position with an implied view buried inside it, and the headline is never the interesting part. The interesting part is the question underneath: what does this person have to believe for this to be the right bet?

So when Chevron signed a twenty-year agreement on June 22 to sell roughly 2.67 gigawatts of behind-the-meter power to a Microsoft data center in Reeves County, and the trade press filed it under “oil major becomes an AI infrastructure company,” the framing was not wrong so much as incurious. Infrastructure is the word you reach for once a bet has paid off and you would prefer everyone forget it was ever a bet. What Chevron actually signed is a portfolio of nested wagers, stacked one inside the next, each with an implied view you can read if you hold it up to the light. There is a bet about power prices and a bet about turbines and a bet about who needs whom more across a twenty-year table. And underneath all of them, doing quiet work almost nobody has bothered to price, is a bet against the Permian itself: against the value of the very molecule that made Chevron rich in this basin in the first place.

That last one is the one we do for a living. Hold that thought. We are going to walk down the whole stack to reach it, because the bet at the bottom only makes sense once you have seen the bets stacked on top of it.

Four Ways to Sell a Molecule, Four Different Things to Believe

Start with the decision Chevron actually faced, because it is the decision every gas-long producer in America is facing right now. A molecule of Permian associated gas can be sold four ways:

1. as gas, to whoever is buying gas;

2. as firm supply, to someone who burns it for power and pays you for the molecules;

3. as merchant electrons, which you make yourself and sell into the grid; or

4. as LNG, chilled and shipped to a buyer indexed to a price on another continent.

These are not styles. They are positions, and each one pencils only if you believe a specific thing about the next ten to twenty years. Suppose you are Chevron, sitting on a molecule that has lately been worth less than nothing at the hub. Which of the four you pick is a confession about where you think that molecule is worth the most, and for how long.

1. The utility PPA, which is the bet Chevron made.

Chevron’s structure, signed through a subsidiary called Energy Forge One LLC, is the one that looks least like an oil and gas decision and is therefore the most revealing. Microsoft pays a fixed capacity payment to reserve the 2.67 gigawatts, and the variable cost, the gas fuel itself, passes straight through to Microsoft. Chevron has, in other words, reverse-engineered itself into a regulated utility: it earns a return on the iron and on the availability, and it has handed the commodity risk to a counterparty rated AAA. Reported economics put the first phase near $7 billion of capital (Chevron has not confirmed the figure, so treat it as reported, not disclosed) against a mid-teens return, first investment decision targeted for late 2026, first power in 2028. What the power actually clears at, I don’t know, and neither does anyone quoting it: broker estimates run from the low $60s per megawatt-hour to north of $85, which is less an estimate than a confession of how little is public. The deal is variously described as throwing off more than a billion dollars of annual EBITDA to the New Energies segment or roughly $250 million of annual free cash flow from 2029. Those are different metrics measured at different points, and the gap between them is the gap between a press release and a model.

Here is the implied view, and it is the whole point. You do not lock twenty years of power economics over your own gas, and then volunteer to pass the commodity through to someone else, if you think the commodity is going to be worth a lot. The PPA is a bet that the ten-year Waha outlook is weak. Chevron looked at the value of its own molecule at the hub, looked at the value of that same molecule converted to a reserved electron behind the meter, and decided the second number was so much larger and so much steadier that it was worth $7 billion and twenty years to stop selling gas and start selling availability. This is, to be clear, a smart trade: a way to be short your own molecule for two decades and collect a utility return for the privilege. It is also about the most bearish thing a Permian operator can do at the wellhead while keeping a straight face. You don’t take it if you think the molecule comes back.

2. Sell it like gas, which is the EQT bet.

Up in Appalachia, EQT looked at the same demand wave and made the opposite wager. Rather than build and own the generation, EQT signed long-term firm supply into power and data-center load: roughly 665 million cubic feet a day to the Homer City redevelopment (a retired Pennsylvania coal site being rebuilt as a multi-gigawatt campus) and about 800 million a day to an industrial park on the Ohio River, something on the order of 1.5 billion cubic feet a day of incremental demand contracted as molecules, priced index-plus. EQT keeps the commodity exposure it is famous for being the low-cost holder of (operating costs around $1.09 per Mcfe will do that), and bets that in-basin demand tightens the very differential that has haunted Appalachia, pulling its realized price up toward the hub instead of paying to ship gas away from it. It is the same demand wave Chevron is reading. EQT just read it as a reason to hold the molecule tighter rather than hand it off.

3. Burn it and sell the power merchant, which is the BKV and Diamondback bet.

A third group is splitting the difference: own the generation, but sell at least some of the output into the merchant market rather than under a fixed twenty-year capacity payment. BKV is buying brownfield combined-cycle plants in Texas at a fraction of greenfield cost (Temple I for about $570 per kilowatt in 2021, Temple II for roughly $460 in 2023, against the $1,000 to $1,200 it costs to build new) and underwriting incremental merchant PPAs at $80 to $100 per megawatt-hour. Diamondback has floated a joint venture with independent power producers to put gas-fired plants on its own Permian surface, pitching cheap land plus cheap gas as the package. The implied view here is the most aggressive of the three. It is the only one of the four that needs ERCOT to stay scared, tight enough for long enough that selling electrons at the margin beats the safety of a fixed contract.

4. Chill it and ship it, which is the Antero and Caturus bet.

The fourth answer ignores domestic power entirely and bets on the spread between American gas and the rest of the world. Antero sends roughly 2.3 billion cubic feet a day, about three-quarters of its production, into the Gulf Coast LNG fairway, and sits on transport that has been clearing a premium to Henry Hub (the Tennessee 500-leg path has run something like sixty-four cents over Henry Hub for calendar 2026, tapering to the high twenties by 2028), while keeping its NGL barrel almost entirely unhedged to catch the export bid. Caturus has gone a step further and is building the export itself: the Kimmeridge-backed South Texas producer (Abu Dhabi’s Mubadala owns 24.1 percent of it) is aiming more than a billion cubic feet a day of Eagle Ford dry gas at Commonwealth LNG, the 9.5-million-tonne, roughly $12.5 billion liquefaction terminal it controls on the Louisiana coast, with about 7 million tonnes a year already sold on twenty-year contracts to international buyers (EQT, Aramco, PETRONAS, Mercuria, Glencore), a final investment decision due in 2026 and first cargoes in 2029. The bet is that a global index beats a domestic one, that the molecule is worth more measured in Rotterdam or Tokyo than anywhere a pipeline can reach it at home. Antero is positioning for the export bid. Caturus is building the thing that sets it, which is a more expensive way of believing the same number.

Four companies, four structures, four incompatible views of the same future, all defensible, all funded. That is what a market under genuine uncertainty looks like: not consensus, but a spread of bets wide enough that somebody is going to be wrong and nobody yet knows who. Which leaves the only question worth asking once the bets are lined up side by side. Under what conditions does each one actually win?

Subscribe now

When Each Bet Wins, and the Range That Decides It

A bet is only as good as the price you got and the world that shows up. So define the world. The parameters that decide which of these four wins are not exotic: the path of Waha and Henry Hub over the term, the power price you can clear, the capital cost per kilowatt, the capacity factor you actually run, and the length of the contract. Move those five and the ranking reshuffles.

The cleanest way to see it is the napkin version: convert one molecule into one electron and price it both ways. A thousand cubic feet of Permian residue gas carries about 1.04 million Btu. Run it through Chevron’s chosen turbines at a design heat rate near 6,300 Btu per kilowatt-hour and you get about 0.1651 megawatt-hours of electricity out the other side. At a $100 PPA that molecule is worth $16.51 as power. At an $80 PPA it is $13.21; at $130 it is $21.46. The full grid of heat rate against tariff runs from roughly $11.89 to $22.53 per Mcf of gross power value, and the entire arbitrage lives in the distance between those numbers and what the same molecule fetches as gas.

This is where the honest range matters, because the gap between the conservative case and the actual 2026 case is enormous, and we are going to carry both rather than quote whichever one flatters the thesis that day.

The conservative case is the one our own model was built on before the year got strange: Waha around $1.50 and a PPA around $65, run through the thirstier bridge-generation heat rate rather than the efficient turbines. Even there the arbitrage is real. Push the conservative inputs through the same conversion and the molecule is worth about eight dollars as power against a dollar and a half as gas, better than five to one, which is already enough to change every downstream decision in this article.

The actual 2026 case is not conservative at all. Waha has spent the year below zero. The hub averaged about negative $2.37 per million Btu early in the year, printed negative $3.30 in May, and stayed underwater for 89 consecutive trading sessions, which obliterated the prior record of 27 set in 2024 (records in this market do not get broken so much as embarrassed). For stretches this spring the Permian was paying people to take its gas, and the prompt month traded as low as negative $5.69. Against a negative Waha the power conversion stops being a re-rating and becomes a rescue. The molecule went from a liability you pay to dispose of to a reserved electron worth sixteen dollars, and the spread is the entire number, not a premium layered on top of one.

Put the two together and you get the range that actually decides the bet. The premium of behind-the-meter power value over the Waha alternative runs somewhere between roughly $13 and $16 per Mcf in the published scenarios, and wider than that on the days Waha is deeply negative. That is the prize. The risk is not in today’s number. It is in the back half of the term, because the same forces that crushed Waha are scheduled to ease: more than 11 billion cubic feet a day of new egress is set to leave the basin through 2030 (Blackcomb at 2.5 Bcf/d late this year, Hugh Brinson’s first phase around the same time, and a queue behind them), and Permian gas production that ran 17.2 Bcf/d in 2021 and 27.6 in 2025 keeps climbing toward 28-plus by 2030. The forward curve already converges toward Henry Hub through the late 2020s, with serious modelers pegging the Waha discount settling into roughly negative 25 cents to negative $1.25 per Mcf once takeaway catches up. So the thing Chevron is betting against is scheduled to partially fix itself, on a known timetable, and Chevron signed anyway, for twenty years, which tells you how long it thinks the broken part lasts.

So read the four bets against that range. The utility PPA wins biggest precisely in the world Chevron is implicitly forecasting: Waha weak-to-negative for years, power firm, the commodity a thing to be rid of. It wins least in the world where egress overshoots and Waha snaps back to a healthy positive number, because then Chevron handed twenty years of commodity upside to a AAA counterparty in exchange for a utility return, which is a fine outcome and a small one. The EQT bet is the mirror image: it needs the basis to tighten, and it loses if demand disappoints and Appalachia stays long. The merchant bet needs ERCOT to stay scarce and is the first to break if it doesn’t. The LNG bet needs the international spread to hold against a wave of global liquefaction. None of them is free. Each is a price on a different probability, and only one of the four has actually shown you its hand, because the structure Chevron chose makes no sense unless Waha stays broken.

The Equipment Is the Position

Here is where the bet stops being a spreadsheet and becomes steel with a lead time, because the monetization view does not just sit in a contract. It gets poured into a foundation, and the equipment you choose is the bet made physical and made expensive to reverse.

Chevron’s deal leans on two kinds of iron. The bulk of the generation comes from GE Vernova heavy-duty gas turbines, the 7HA class, running at roughly 60 percent thermal efficiency and that 6,300 Btu per kilowatt-hour design heat rate, with Caterpillar’s Solar Turbines supplying additional dispatchable capacity. The partner structure is its own tell: Engine No. 1, the activist-turned-energy-investor, launched a company called Joulent (built with GE Vernova) that holds a 50 percent option to co-fund the project. That option is not a footnote. It is optionality in the literal finance sense, a way to commit to the slot without committing all of the capital, which is exactly what you do when you like a bet but would rather keep your downside sized to your conviction.

The heavy-duty turbine is a bet on efficiency at baseload. It sips gas relative to the alternatives, which matters enormously when you intend to run it flat-out for twenty years and every hundredth of a point of heat rate compounds across two decades of fuel. The competing iron is the reciprocating engine, the Caterpillar 3600 class and its kin, and it makes the opposite bet. A recip is less efficient (heat rates of 7.5 to 9.0 against the turbine’s 6.3, an efficiency of maybe 40 to 52 percent), and it costs less up front (call it $1,700 to $2,000 per kilowatt installed against $2,500 to $3,000 for combined cycle), and it does one thing the turbine cannot: it starts almost instantly, full power in under five minutes, and it barely loses efficiency at part load. For a workload that swings hard and fast, the recip’s responsiveness is worth paying inefficiency for. For a hyperscale baseload that wants to run pinned at the top of its range, the turbine’s efficiency is worth paying for. Chevron is building for steady AI baseload over twenty years, so it bought efficiency, and paid for it in capex it cannot easily walk back.

And the choice is getting harder to make and harder to unmake, because the iron itself is now the scarce asset. GE Vernova is sitting on something like 100 gigawatts of gas-power commitments (roughly 44 gigawatts of firm backlog plus another 56 in slot-reservation agreements) inside a corporate backlog north of $160 billion. Caterpillar’s backlog hit a record $62.7 billion, with its reciprocating-engine book up more than three and a half times since the start of 2024 and lead times past 200 weeks, which is to say you can order the engine today and take delivery sometime after the contract you bought it to serve has begun. The top turbine makers cannot deliver a new heavy-duty machine before the end of the decade. Chevron locked seven GE Vernova turbine slots with deliveries in late 2026 and 2027, and those reservation agreements carry non-refundable down payments of 10 to 20 percent of the equipment cost. Pause on that. The slot reservation is itself a bet, a non-refundable premium paid for the mere right to build, in a market where the right to build is suddenly the binding constraint. You are no longer just betting on power prices. You are betting on your place in a four-year queue, with real money you don’t get back if you are wrong about needing the place.

Every bet so far has been visible from the outside. The PPA is in a press release. The turbine order is in a backlog. The Waha strip is on a screen. They are the bets everyone can see, which is precisely why everyone is writing about them. The bet nobody is pricing is sitting in a field in Reeves County, burning gas to do its job, and it happens to be the one we can actually measure.

A Third to a Half of the Fleet Already Stands Behind It

We have spent years building the one thing nobody outside a compression vendor has ever had: an independent census of the iron. Who runs which units, where, of what make and vintage, behind which operator, across the Permian, reconciled against what the vendors themselves disclose. We are going to use it here at the level the argument needs and hold the unit-by-unit roster back, because the roster is the product and the argument is the advertisement. What the argument needs is the shape of Chevron’s fleet, and the shape is unusually clean.

Chevron operates ~100 active compression units in the Permian, 132,820 horsepower, and the fleet is almost entirely one make. Caterpillar is 94.6 percent of the horsepower. Two engine families do most of the work: the G3608, twenty units at 50,750 horsepower, and the G3606, twenty-six units at 48,850 horsepower, which is forty-six big central-station machines carrying 75 percent of every horse Chevron runs in the basin. The largest iron lives in New Mexico, where Eddy and Lea counties hold 48 percent of the basin horsepower on thirty units, at stations whose names sit in the public air permits: Sand Dunes, Dagger Lake, Antelope, Pronghorn, Hayhurst, Salado Draw. And roughly half of all that horsepower, about 51 percent, runs under a single compression provider, which makes one packager the dominant landlord on Chevron’s Permian iron (we know which one, and we can age every unit it runs). Hold that fact too. It comes back.

Now lay the data center on top of the fleet. Delivering the gas that feeds 2.67 gigawatts takes somewhere between a third and a half of this iron. Size it against the efficient turbines Chevron actually bought and it is about 46,000 horsepower; size it against thirstier bridge generation and it is closer to 62,500; either way it is twenty to twenty-nine of those big G3606 and G3608 stations, the Sand Dunes and Dagger Lake and Antelope machines, give or take. Read that sentence twice, because it is doing something the deal coverage missed entirely. A third to a half of Chevron’s Permian compression fleet is now standing behind a twenty-year power contract with Microsoft. Nothing in the press release, the broker notes, or the deal’s own org chart so much as mentions a compressor, which is roughly the standard treatment for the asset that actually moves the gas.

And the database ages the iron. The build dates span a quarter century, from a lone 1999 unit to a single new 2024 machine at Antelope, and the horsepower that matters here sits in two clusters: a wave of large central frames built between 2008 and 2010 and a buildout of smaller wellhead units across 2013 to 2015. Date the big G3606 and G3608 stations that feed the data center and they land in that 2008-to-2010 window, iron sixteen to eighteen years into a service life, running lean-burn technology two engine generations behind what you would install new today. Old, thirsty iron, sitting in the gas path of a twenty-year power contract. For fifteen years that was a deferred-capital problem nobody could justify fixing. Whether it still is depends entirely on what the gas it burns is now worth, which is the arithmetic in the next section.

The Gas the Compressor Burns Was Repriced by a Contract It Never Signed

A bet this size does not stop in the boardroom. It travels all the way down to the person turning the wrenches, and it quietly changes what their job is for. Take a production engineer on these assets. The week is a priority stack: keep the wells flowing, keep the compressors up, chase the next tie-in, hold lease operating expense down. Fuel gas has never been on that stack, and it never should have been. A compressor burns field gas to run (the driver on a G3606 or G3608 is itself a Caterpillar engine, pulling gas off the stream to turn the crank), and across the operated fleet that burn is about 16.7 million cubic feet a day, roughly 1.5 percent of the gas. Priced at the wellhead, it is a rounding error every engineer correctly ignores. You do not optimize a molecule worth a dollar and a half when the well next door makes ten thousand dollars a day. Ignoring it was the correct arithmetic, and it stayed correct for years.

And lately the molecule has been worth less than nothing. At a negative Waha the gas a compressor burns is, strictly, a molecule you were going to pay to dispose of, combusted for free to do useful work, which is the one silver lining of negative prices nobody puts on a slide. So the fuel-efficiency upgrade has never penciled. Every compression engineer in the Permian knows the old iron is thirsty, and every one of them has watched that upgrade die in the capital committee, because the thing it saved was never worth the capital it took to save it.

Priced at the wellhead. That is the assumption the data center just broke.

Now run the same molecule through the new contract. Once the gas this fleet moves is feeding 2.67 gigawatts of reserved power, the molecule a compressor burns is no longer worth the Waha price. It is worth what it would have made as power, the PPA price divided by the plant heat rate, on the order of sixteen dollars per million Btu at a hundred-dollar tariff. The compressor’s fuel gas did not change. The molecule did not change. A contract it never signed changed what the molecule is worth, by roughly an order of magnitude, and in doing so it lifted fuel-gas efficiency from the bottom of that production engineer’s priority stack to somewhere near the top. The strategic bet, made once at the top of the chain, reaches down and rewrites what the person at the bottom is paid to optimize.

Put the arithmetic on the table, both ways, because the gap between the two is the finding. Repower the old, thirsty iron (call it the 60,000 horsepower of pre-2015 large frames that sit in the data center’s gas path) with modern lean-burn engines, and you cut fuel consumption on that horsepower by about 12 percent. That frees up roughly 0.94 million cubic feet a day of gas, about 343,000 million Btu a year, that used to go up the stack and now does not.

Measured the old way, against Waha, that saved gas is worth about half a million dollars a year, and in a negative-price stretch it is worth nothing or less. The project dies, the way it always died.

Measured against the PPA, the same saved gas is worth about $2 million a year in our conservative world ($65 power, the thirstier heat rate, Waha at a buck and a half) and closer to $6 million a year in the world that actually showed up ($100 power, the efficient turbines, a Waha that has been deeply negative). The capex is identical and the molecules are identical. The only thing that moved is where the saved gas goes, and where it goes is now worth four to twelve times what the hub would have paid for it. Over the twenty-year term of the contract, discounted, that single repower is worth somewhere between roughly $18 million and $53 million of value created out of the identical physical project, which is what turns a fuel-efficiency upgrade from an O&M line the capital committee rejects into a revenue-backed investment it should be fighting to fund.

And that is the floor, because it assumes you swap like for like, which no one who has read this far would. The fleet was sized for a job it is no longer only doing. Right-size the horsepower to the throughput it actually moves, optimize the operating points instead of running big frames at part load, and the savings compound on top of the vintage swap. Run the logic across the full hundred-thousand-odd horsepower of G3606 and G3608 iron and the saved gas roughly doubles, the annual value runs to $12 or $13 million, and the twenty-year figure crosses $100 million, all of it out of compression nobody outside the vendor had bothered to age.

There is a twist in here worth saying plainly, because it inverts the instinct most people walk in with. You would think a more efficient power plant makes the upstream fuel savings matter less. It is the opposite. The better the plant’s heat rate, the more electricity each saved molecule becomes, and so the more each saved molecule is worth. Chevron buying the efficient turbines made the compression prize bigger, not smaller. (I had this backwards the first time I ran it, which is how I know it is counterintuitive.) The bets are stacked, and they compound in the same direction.

The bet also changes when you should move. Compression iron goes through major overhauls on a schedule regardless of any of this, and a repower folded into that cycle, or into the data center’s own multi-year ramp toward first power in 2028, lands as a planned swap rather than a capital shock. Do it proactively and the upgrade rides existing maintenance spend with no large step-up in operating cost. Do it reactively, after a failure or a scramble to feed a contract that is already signed, and you pay the premium twice, once for the iron and once for the hurry. So the move is to do it early, on the maintenance cycle, before the signed contract turns a planned swap into an emergency.

Now read the same case from the other side of the contract, because roughly half of this fleet is not Chevron’s iron at all. It is the provider’s, rented under contract, and the engine that just got repriced sits on the packager’s balance sheet, not the operator’s. A compression company that understands this before its customer does is not looking at a fuel-efficiency problem. It is looking at a go-to-market: approach the operator first, fund the new iron, and bundle the upgrade into a longer contract term, which is exactly the trade the equity market rewards a packager for making. The REIT model is unforgiving here, because a packager’s multiple is a function of contracted backlog and duration, so an upgrade the customer already wants, written into a longer term, is the rare deal that improves the asset and the contract at the same time. Get to that number before your competitor and you write the longer renewal yourself, which is the whole game; get there late and you inherit a contract where the upgrade was already funded on someone else’s balance sheet.

This is the part no press release can show you. One strategic bet, made once, at the top of the chain, reaches down and rewrites a production engineer’s priority stack, a capital committee’s hurdle math, and a packager’s sales motion, all at once, and it does it differently for every player depending on the bet they made and the iron they hold. The same map that reads Chevron’s implied view also tells you where the next fleet upgrade, the next contract extension, and the next exposed operator are going to surface. Reading how each bet wins and loses is how you get to plan around the people making them instead of reacting after they have moved.

Resulting in Reeves County

Strip the deal back to what it is and it is one directional bet, made by one operator, that the molecule is worth more as a reserved electron than as Permian gas. A bet that size does not stay where you place it. We have just followed it from a PPA to a turbine slot to the fuel a compressor burns to the renewal between an operator and the landlord that owns its iron, repricing each rung on the way down, and the only people who can see the whole path today are the ones who can read the fleet.

So the discipline is the one Duke prescribed: refuse to result. Read the implied view, not the outcome that has not arrived. Chevron’s structure wins only if Waha stays broken for a decade, which is a real and well-made call, and also one that can be wrong, with a probability attached and a way to lose, which is everything the word “infrastructure” is built to paper over. Everyone downstream is holding a piece of that same wager now, most of them without having chosen it. The operator that reruns its fuel math moves its repowers early. The packager that sees the renewal coming gets there before its rival does. The investor reads the iron instead of the press release.

Whether Chevron is right about Waha, I don’t know. Neither does Chevron; that is what makes it a bet rather than a fact. What I do know is that the bet traveled four steps down a chain the deal never named and came to rest on a compression fleet nobody told it was holding a position. The fleet is real, it is old, and it is sitting in Reeves County with its build dates. Anyway, that is where I would start reading.

Share

Subscribe now

Disclaimer: This article is for informational purposes only and does not constitute investment advice, a recommendation, or a solicitation to buy or sell any security. Kalibr Partners is an independent firm and is not affiliated with, endorsed by, or compensated by any company named herein. The analysis draws on public sources (SEC filings, earnings calls, press releases, and state air-permit and production records) together with Kalibr’s proprietary compression dataset; figures are presented in good faith without warranty as to accuracy or completeness, and certain figures are estimates or reported (not officially disclosed) and are flagged as such. All trademarks and company names are the property of their respective owners.

Subscribe now

When you buy an apartment building, the seller hands you a rent roll. It is a single sheet, and it is the most important sheet in the transaction: every unit, who lives in it, what they pay, when their lease ends. It is also, in the hands of a motivated seller, the most creative sheet in the transaction. Occupancy gets rounded up. A tenant who is sixty days late is still, technically, a tenant. A lease that expires next quarter looks exactly like a lease that runs four more years, because both are just a date in a column, and the column does not tell you which dates the seller is praying about.

So nobody who does this for a living believes the rent roll. They verify it. They send every tenant an estoppel certificate, a boring legal form that asks the tenant to confirm, in writing, the one thing the seller has every incentive to shade: what they actually pay and when they actually leave. The estoppel is a verification technology. It exists because the asset is observable (you can walk the building) but the income is a story (you cannot walk the leases), and in 2022 and 2023 a great many people who had skipped the estoppel and trusted the pro forma discovered the difference between the two when their floating-rate debt repriced and the collected rent turned out to be a number the spreadsheet had been optimistic about.

There is a name for the general problem, and it is older than the apartment syndication. In 1970 George Akerlof published “The Market for Lemons,” and the insight that won him a Nobel was this: when the seller can see the quality and the buyer cannot, the buyer rationally stops believing the seller’s description and prices the average instead. The good used car and the lemon sell for the same number, because from the buyer’s seat they emit the same signal. The only thing that breaks the pooling is verification: a way for the good car to prove it is good, which is to say, a CARFAX. An independent record of the asset that the seller does not author.

Compression is a rental business that runs on a rent roll nobody outside the vendor has ever read.

I have written before that the compression vendor is a REIT. The model is exact enough to use, not cute: the vendor owns the iron, finances it, and rents it to you at a rate that behaves like a yield on book value. A vendor’s accountant would tell you the rate is contract revenue and the book value is on a different schedule, and that is true; the point is not the ledger, it is that the rate moves with the capital base the way rent moves with a building, and reading it that way predicts what the vendor will and will not give you. Kodiak, Archrock, and USA Compression are the three big public landlords of the Permian, and like any landlord they tell a story about their buildings. They tell it to equity analysts on quarterly calls, and they tell a compressed version of it to you across the renewal table, and it is a good story: diversified tenants, blue-chip credit, ninety-eight percent uptime, seven-to-ten-year leases. You cannot verify any of it from your pad. You can see your unit. You cannot see their rent roll.

So we built the estoppel. For the entire Permian.

Before the rest of this, the punchline, because withholding it would be its own kind of story: we reconstructed the Permian compression fleet from the asset up, unit by unit, operator by operator, and then we did the one thing the equity story never invites, which is reconcile it against each vendor’s own Q1 2026 filings. Most of it ties out. That is the point. The bottom-up iron agrees with the top-down disclosures often enough that you can trust the places where it does not, and it stops agreeing in two specific, expensive places, and it sees a third thing the disclosures were never built to show. One landlord’s diversification is a single tenant. All three landlords’ “long-term” lease books reprice a lot sooner than the word “long-term” is doing on the call. And underneath the brand on every skid is a specific engine, with a specific lead time and a specific scarcity, which is the thing that actually sets your leverage and the one thing no rate card will ever show you.

And it is urgent, not merely interesting. The Permian’s gas problem is structural and getting worse: as the basin’s wells age, the gas-to-oil ratio climbs, associated gas keeps rising whether or not anyone wants more gas, and all of it has to be compressed to move. Demand for horsepower here is not cyclical, it is geological. And the iron to meet it sits on a 110-to-180-week lead time for large new builds, which means the installed fleet is not a commodity you can re-source on a phone call. For two and three years out, it is the only fleet there is. When supply is that tight and that slow, every renewal is a seller’s market, and the one thing that arms the buyer is knowing the seller’s book better than the seller expects you to.

Kalibr packages what follows two ways, one per reader. If you run compression spend for an operator, there is a Permian Compression Benchmark built for your side of the table. If you sell compression, or finance the people who do, there is a Permian GTM Targeting read built for yours. A fair question is how one shop arms both sides, and the answer is the thing that makes the data worth trusting: the iron is the iron, and the fleet does not rearrange itself depending on who is looking. The dataset is neutral; the representation is not. In any single engagement we sit on one side of the table and say which, and the read built for the other side is not one we run against the client we took. Both products are below. If either is your problem, the fastest path is to message us; the rest of this explains why the data underneath them is worth the conversation.

The Iron Agrees, Until It Doesn’t

Before the rent roll, the building. How big it is, and why you should trust a number we built from the iron up instead of one the vendor handed you, which matters more than it sounds: the entire argument that follows depends on the bottom-up agreeing with the vendors, right up until the moment it doesn’t.

The Permian carries roughly 6,380 compression units in our database, and we work the 2,855 we can attribute to one of the three public landlords by name, 2,360 of them active (1,765 in Texas, 1,090 in New Mexico). Kodiak, Archrock, and USA Compression are the consolidated, investable core of this market, and the three counterparties an operator is most likely to be sitting across from when the lease comes due; the rest of the basin is a fragmented tail of smaller fleets that this analysis is not about. Every number in this piece is a unit we can put a vendor’s name to, and the totals reconcile to the vendors’ own filings. If we cannot attribute it, we do not report it.

Here is the first thing the iron says, and it says it in agreement with the vendors, which is exactly why it is useful. Every one of these three is bigger in the Permian than it is anywhere else. Kodiak’s company-wide fleet averages 977 horsepower per unit, Archrock’s around 900, USA Compression’s 694. Their Permian fleets average 1,411, 1,421, and 1,219. The basin is where the large iron lives, the G3600s and the high-rate gathering units, the equipment with the highest rate per horsepower and the longest lead time. When Kodiak’s filings say its fleet is roughly 80% large horsepower above 1,000 HP, our Permian cut puts it at 80.2%. When the companies rank themselves by fleet age, our vintage data puts them in the same order they put themselves. The bottom-up reconciles to the top-down.

That reconciliation is the credibility engine, so I want to be honest about where the data argues “for” the vendors before it argues against them. Kodiak really does run the youngest fleet of the three. USA Compression really is the most pressured, and its own disclosures say so as loudly as our data does. Archrock’s equipment really does sit on location for years, exactly as it claims. A model that only ever found problems would be a model worth ignoring. This one finds agreement first, which is the only reason the disagreements are worth your time.

Now the concentration. Run a Herfindahl on each vendor’s Permian operator base and Kodiak scores 1,351 against Archrock and USA Compression at 968 apiece. That gap is the whole story of the next section, and it runs the opposite direction from the one the equity narrative is selling.

A Third of Kodiak Is One Customer

Kodiak is the market’s favorite compression name, and the case is good. Youngest fleet, highest rate per horsepower, 98% utilization, a pricing environment management expects to run “to 2027 and beyond.” The customer base is described the way you would describe a well-run REIT: top ten customers more than half of contract services revenue, investment grade above 60%, a disclosed roster that reads like a blue-chip index, supermajors and the largest shale independents and the midstream majors. The company describes its mix as roughly 30% upstream and 70% midstream. The word the story wants you to hear is “diversified”.

Here is the rent roll.

Company A, a single investment-grade supermajor, is 420 of Kodiak’s Permian units. That is 31.7% of its attributed Permian fleet by count and 33.1% by horsepower: 613,790 of 1,856,266 horsepower, set behind one logo. The top three operators are 50.7% of the units and 54.7% of the horsepower. And this is not cheap gas-lift volume parked to pad a count. Company A’s Kodiak units average 1,461 horsepower, with 81 of them at 1,500 or above. It is the largest iron in the book, behind the single largest counterparty.

Two honest notes on basis before the conclusion, because they are the difference between a finding and a number you should not trust. The disclosure counts revenue; we count units and horsepower, which is not the same denominator, and we will not pretend it is. And our share is the operator we attribute to each skid from asset data, not a contract we have read. Neither caveat moves the shape. Operator attribution is the part of this dataset we stand behind hardest, populated on essentially every unit, and a third of the core-basin iron sitting behind one counterparty is a fact about the fleet, which is the thing we can see directly, where revenue is the thing the disclosure chooses to show you.

“Top ten customers more than 50% of revenue” is true. It is also the sentence doing the most work in the whole disclosure, because it is built to make ten sound like the operative number when the operative number is one. A third of the core-basin fleet behind a single counterparty is not what diversification means. It is what a levered bet on one drilling program means, and the word for the risk it creates is not concentration, it is hold-up: when you have sunk capital that is specific to one counterparty (and a 1,461-horsepower unit on Company A’s pad is about as specific as capital gets), the counterparty holds the option, and everyone in the relationship knows it.

It gets more concentrated, not less, which is the part the growth story inverts. Kodiak is deploying roughly 150,000 to 170,000 horsepower of new builds in 2026, more than 90% of it Permian, essentially all pre-contracted, and the equity story reads that as broad secular growth. The sets say it is feeding the whale. Company A was 46 of 68 Kodiak Permian sets in Q2 2025, 22 of 62 in Q3, 67 of 98 in Q4, 35 of 66 in Q1 2026. Call it a third to two-thirds of new sets every quarter, going to one operator. The growth capex is not broadening the customer base in the Permian. It is deepening the single name.

None of this makes the bet a mistake. Anchoring your fleet to one of the best-capitalized operator in the basin is what you would do if you could: Company A pays, Company A drills through cycles, Company A is the counterparty you want when oil prices sell off. The bet is the right bet. It is also, unmistakably, a bet, with a name and a risk profile, and the equity story has dressed it as the absence of one. For the operator on the other side of it, the gap between those two readings is the whole opportunity: the most important fact about your Kodiak relationship may be how badly Kodiak needs Company A’s drill schedule to stay exactly as good as it is.

Compare the other two landlords, because the contrast is the proof. Archrock’s largest Permian customer is 144 units averaging 932 horsepower, with 4.2% of them at 1,500 or above. USA Compression’s largest is around 100 units averaging 713 horsepower, 5.3% large. Their whales are small wellhead fleets. Kodiak is the only one of the three whose biggest tenant gets the biggest iron, which is the asset-level signature of a concentrated bet that the other two, whatever their other problems, did not make. Hold that fact, because the build years say it again in the next section, by a different measure entirely.

Subscribe now

Read the rest. The engine map, the roll-off calendar, and the operator’s playbook are for paid subscribers.

Read more

Subscribe now

In August 2004, British Airways added a fuel surcharge to its long-haul tickets. It was £5 per round trip, which is to say it was nothing, which is the correct size for a brand-new line item: small enough that nobody calls, large enough to exist. And it was hard to argue with. Oil prices were climbing, jet fuel is refined from oil, airlines buy oceans of jet fuel. This was not a price increase, went the framing. It was a pass-through. A regrettable, temporary, externally imposed pass-through.

By early 2006 the pass-through was £60.

Now, oil really did get more expensive over those eighteen months. That part was true, publicly visible, and repeated on every evening news broadcast in Britain, which was precisely the point. The headline did the persuasive work. A passenger could see that oil was up. What the passenger could not see was how much oil actually flows into the cost of one seat on one 747 to New York, which meant the size of the surcharge was, from the customer’s side of the counter, unverifiable. £5 felt plausible. So did £20. So, apparently, did £60. The number was tethered to nothing except what the story would bear, and the story, it turned out, would bear a 12x increase.

Also, the two airlines were on the phone with each other. I should mention that.

Between August 2004 and January 2006, on at least six occasions, British Airways and Virgin Atlantic informed each other about planned surcharge changes instead of setting them independently. You are not supposed to do that. It is, in fact, extremely illegal. And the scheme ended the way these schemes always end, which is not because customers got smarter. Customers couldn’t get smarter; they had no verification technology, and you cannot out-read a number that means whatever its author says it means. It ended because Virgin walked into the UK Office of Fair Trading in 2006 and confessed, in exchange for full immunity. In August 2007 BA was fined £121.5 million by the OFT (a record at the time, later reduced to £58.5 million on appeal in 2012, which is the kind of detail that matters to BA and to nobody else) and pleaded guilty in the United States to a criminal price-fixing charge carrying a $300 million fine. In 2008, BA and Virgin agreed to refund roughly $210 million to about 7.2 million passengers. One third of the surcharges they had paid. The other two thirds had, presumably, gone to fuel.

I want to be careful about what the crime was, because the careful version is the useful one. Charging more for fuel when fuel is expensive is legal, ordinary, and arguably honest. The fuel surcharge was never the crime. The crime was coordination: the two carriers removed the one mechanism (a competitor willing to undercut an inflated number) that could have checked the number. BA’s chief executive said afterward that passengers “had not been overcharged.”

Sit with that sentence. It is a perfect sentence. If a surcharge means whatever the airline says it means, then no passenger can ever be overcharged, by construction. The defense is the business model.

There is a name for all of this. In 1982, two economists, Vincent Crawford and Joel Sobel, published a paper in Econometrica called “Strategic Information Transmission,” and it became the founding document of what game theorists call cheap talk: communication that is costless, non-binding, and unverifiable. Here is the game. One player knows the true state of the world, say, what its costs actually did. The other player cannot observe the state, hears a message, and writes a check. The catch is a single parameter called bias: the informed player’s preferred outcome sits a little above the truth, in its own favor, always. The vendor does not want you to pay what costs justify. It wants you to pay that, plus something.

You can probably already see the problem, and Crawford and Sobel proved you cannot see your way out of it. Two of their results are worth carrying around for the rest of your career. First: when the sender benefits from your response, there is no equilibrium in which precise claims can be believed. None. Not because salespeople are liars, but because if exact numbers were believed, exact numbers would be inflated, and a rational listener prices the inflation in before the meeting starts. This is the part people miss: it is not a theory of lying. It is worse than that. It is a proof that even a buyer who knows the game cold cannot extract truth from words, because the words are the only thing being offered and the words are free. Second: the bigger the sender’s stake in your overreaction, the less its messages can carry. Communication degrades from numbers into vibes (”costs are up a lot”), and at high enough bias into nothing at all, a state economists call the babbling equilibrium, and your grandmother called the boy who cried wolf, and you may recognize as certain quarterly business reviews.

So: one way to read the BA story is as an antitrust case. Another way to read it is as a 30,000-foot experimental confirmation of a 1982 theorem, in which an unverifiable cost claim, bolted to a true and visible headline, drifted from £5 to £60 because nothing tethered the message to the state. The most expensive thing on the ticket was not fuel. It was the story about fuel.

As part of my job, clients often call with decks they receive that carry a very good story.

Before I tell it, the punchline, because withholding it would be its own kind of cheap talk: we have now run the sequence that follows for a string of clients, against a string of vendors, conflict decks and tariff decks alike, and the surcharge it is built to answer almost always ends at the same number, which is zero. The cumulative figure we have taken off these decks has a couple of commas in it. What follows is how, and why it keeps working.

The Market Designed to Feel Unwinnable

Before the deck, the envelope. I am going to tell you how much money is in it, and then I am going to tell you why almost nobody you employ can verify the number on a single tote of it.

Let’s start with every Production Engineers favorite cost center; chemicals.

North American E&P operators spend roughly $11.5 billion a year on oilfield chemicals. Call it $2.0 to $2.3 billion of production chemicals (the corrosion inhibitors, scale inhibitors, biocides, H2S scavengers, demulsifiers, and paraffin treatments dripping into your wells right now), with the rest in drilling fluids and stimulation chemistry. This is not an abstraction that lives in a market study. Kalibr benchmarking for a public mid-market Permian operator runs about $0.71 per BOE in chemicals and treating, roughly 6.5% of its LOE. Frac chemistry runs 5 to 10% of horizontal well D&C, in a world where the completion is 54% of the well. Corrosion and scale inhibitors are typically 15 to 20% of total chemical spend. Each.

A real line item, then. Here is the uncomfortable part: it is the worst-audited real line item you have, and it is built that way.

Chemical contracts split roughly 50/50 between the chemistry and the service wrapped around it, the field techs and the monitoring and the formulation knowledge, which means the thing on your invoice is not a commodity with a price. It is a bundle with a narrative. Two suppliers control roughly 40 to 50% of US land production chemistry. Sole-sourcing is the norm, because trialing a new chemistry on producing wells feels like performing surgery to save on bandages. And the slack is not hypothetical: simply auditing an active chemical program yields 5 to 10% in direct savings for the average operator, and automated dosing routinely cuts consumption 15 to 25% by adjusting injection rates more often than a human with a route schedule can. Do the Feynman check on that pair of numbers. If an audit alone finds 5 to 10%, the program was never within a percent or two of right. It was loose by double digits, for years, and the money was simply going. Quietly. With excellent service.

Why does nobody look? Because production chemicals are psychologically miserable to own. You are paying to prevent failures that have not happened, which means the program’s success is indistinguishable from its excess. The internal dialogue runs something like:

Year one: Nothing failed. Are we overtreating? Cut the program.

Year two: An ESP died of scale. Why did we cut the program? Double the program.

Both of those are emotions. Neither is a model. Add switching costs (field trials, residual chemistry risk, a vendor tech who knows your wells better than your own foreman does) and you get a buyer who cannot verify, cannot easily leave, and negotiates from whichever failure is freshest.

Now run the Crawford and Sobel checklist. Every trait above does one of two things: raises the seller’s bias, or lowers the buyer’s ability to verify. Production chemicals are not a market where cheap talk occasionally happens. They are the natural habitat. The only missing ingredient is a headline.

The Part Where the Headline Shows Up on an Invoice

Every few quarters, the world provides one. A tariff schedule. A hurricane in the Gulf. A canal blockage, a port strike, a regional war. The genre is durable because the ingredients are always in stock: a disruption that is real, visible, and quantified daily in the press, attached to a cost claim that is none of those things. You have seen the catalog. The pandemic logistics rider. The tariff adder. The force majeure letter that somehow never arrives when prices fall.

The one that landed on our client’s desk cited a geopolitical conflict, and the supporting evidence was genuinely strong, the way rising oil prices in 2004 were genuinely strong. Feedstock markets convulsed. Chemical majors announced double-digit increases. A global petrochemical index posted its steepest one-month move in decades. None of that was in dispute. Headlines rarely are. That is what makes them load-bearing.

Days into the news cycle, a major service company sent our client a deck. It was a good deck. I want to be fair to it, because it was genuinely professional work: clean charts of crude against its 30-day rolling average, freight indices by trade lane (you know a deck is serious when the freight indices come out), a tidy stack-up of cost categories building to a total impact. And then the proposal: a 12% surcharge on every invoice line item, effective in two weeks, tiered to the 30-day rolling average of a benchmark crude index, with the reassurance that there would be no changes to the price book.

One way to read no changes to the price book is as restraint: we are not repricing you, we are merely surviving the news together. Another way to read it is that a surcharge that never enters the price book never has to leave it, compounds with every gallon of volume growth, and appears in no benchmark anyone will ever compile. The crude tie reads as objectivity, and it is actually an anchor swap: production chemicals are not made of crude oil, but crude is the one index the customer already watches every day and cannot argue with. And the ratchet was asymmetric: up with the index immediately, down on an evaluation schedule, and below the baseline tier the operator does not receive a credit, the surcharge simply rests. Heads, pass-through. Tails, also pass-through.

It is, structurally, a fuel surcharge. The headline is visible. The exposure is not. The number means whatever its author says it means.

Unless you build the thing BA’s passengers never had.

We (Kalibr) have built it. The principle is not exotic: a production chemical is a recipe, and recipes can be costed. Decompose each product family into its feedstock families, weight each feedstock by what the formulation actually contains, price each against a public index, and let only the layer of cost that is actually made of molecules move with the molecule markets. Then perform the one operation the vendor’s deck conspicuously declined to perform, which is multiplication.

Here is the whole trick, with round numbers you can check on a napkin. One feedstock spikes 20%. It is 30% of the raw material basket. Raw materials are half the cost of the finished product. The product impact is 0.20 × 0.30 × 0.50, which is 3%. Not 20%. The deck quoted the 20% and applied it to products as if they were made of nothing but the headline, which is the compositional sleight of hand the entire genre depends on, and it dissolves on contact with three multiplications a junior analyst can do in the margin of the deck itself.

And even the napkin is generous, because it assumes everyone is paying spot. You probably are not, and your supplier almost certainly is not. The large formulators buy their feedstocks the way you wish you bought theirs: forecast far in advance, contracted long, buffered with bulk pre-buys. Innospec, a publicly traded oilfield chemical specialist with the same feedstock exposure, says it plainly in its 10-K:

“We use long-term contracts (generally with fixed or formula-based costs) and advance bulk purchases to help ensure availability and continuity of supply, and to manage the risk of cost increases.”

That is a supplier explaining, in a federal filing, that its input costs are built not to move when the headline does. So when the surcharge letter cites the spot market, the right question is not whether spot spiked. It did; you can read. The right question is how much of the seller’s actual cost basis touches spot at all, and the honest answer is usually: a sliver, with a lag measured in quarters.

I can hear the objection forming. What about force majeure? What if the supplier’s own contracts broke? This is the lovely thing about negotiating with publicly traded counterparties: disclosure is not optional for them. A disruption that materially threatens a supplier’s cost structure or supply chain has to show up in its filings. I will give you the honest calibration, though, because the risk-factor section alone will not settle it (those pages are written by lawyers to be maximally broad, and they warn about everything, always). Read where the numbers live instead. Segment margins. Management’s commentary on cost trends, which must be current. Inventory on the 10-Q, because a supplier sitting on raw materials bought before the spike is delivering you product made at the old cost basis, whatever the cover letter says. An afternoon of financial due diligence parses a force majeure story with reasonable conviction, and a vendor whose own filings show expanding margins and fat pre-bought inventory is telling a story that does not survive its own paperwork.

One more thing, because the worst version of this deserves its own paragraph: the surcharge that arrives mid-term, on a contract with agreed pricing. Allow me a brief soapbox. Term is one of the most valuable things you can give a vendor. It is revenue certainty in a spot-leaning industry, and the market pays handsomely for revenue certainty (listen to any compression company earnings call and count the minutes management spends advertising contract duration to the street). So when you granted term, you were not receiving a discount as a favor. You bought cost certainty and you paid for it with revenue certainty. That was the trade. A surcharge in the middle of it is not a price adjustment; it is a re-trade, and ours, of all industries, should recognize one on sight. We are professional makers of probabilistic, risk-based decisions. We do not drill a well, watch it come in under type curve, and send the service companies a letter requesting a retroactive discount because our risk assessment aged poorly. The vendor’s procurement bet aging poorly is the same event, wearing their jacket. Holding the line on a term contract is not hardball. It is the contract.

Back to our deck. Run that arithmetic across the client’s actual portfolio and the event’s defensible impact came out to roughly 7% weighting product families equally, closer to 6% weighted by spend. Against a flat 12% claim, call it a 2x overstatement on the typical family. And the texture underneath is what made it unanswerable. The crystal-modifier polymers dominating one family’s cost had not moved at all (the claim there was overstated roughly 3x). And one family, hydrate inhibitors, which are essentially bulk methanol in a branded tote, justified *more* than the surcharge asked for, because the disruption had taken a real slice of global methanol capacity offline. We led with that one. When your model occasionally argues for the other side, people start believing your model. (This is also a reasonable test of a consultant, but that is a different newsletter.)

This is the regime change, and the game theory language is worth the thirty seconds because it tells you the model is not a tactic. Crawford and Sobel proved you cannot fix cheap talk by listening harder. A parallel literature (Grossman and Milgrom, 1981) proved what happens when claims become provable: the equilibrium flips. Exaggeration becomes detectable, detection becomes expensive for the exaggerator, and silence itself becomes informative, because a vendor with a genuine cost case will sprint to show you receipts. A should-cost model does not improve your position in the old game. It deletes the old game and replaces it with one where “our costs went up” stops being a price and starts being a hypothesis.

The minimum viable version, for anyone who owns chemical LOE, is honestly not much: a one-page decomposition of your top product families against four or five public indices, refreshed quarterly. The full version is contracts indexed to feedstocks at margins both sides accept with symmetric movement, and an auto-RFP that quietly re-shops your program when the indices fall and your contract does not. (A vendor who objects to symmetric indexing is telling you, with admirable economy, what the last surcharge was for.) That is the layer Kalibr built under compression, now running under chemical spend. The verification technology, permanently on.

Chemicals is not special. It is simply the category I was standing in when the deck arrived, and, if anything, the easy one. It is the only line we track that is a pure commodity, which is why it decomposes as cleanly as it just did. A chemical is a recipe; there are no performance trends riding on top of the molecules, no KPI story to pull out of the price. The others are harder. Frac, OCTG, and compression each carry a service-and-performance layer, a supply-and-demand layer that moves price independent of cost, and a financial decoupling you have to run before “should-cost” means anything at all. We build that for each of them, because the discipline does not change even when the anatomy does: separate the part of the price that is a real, indexable commodity from the part that is a story, tie the first to a public number, hold the second to ordinary inflation, and write symmetric movement into the contract. The surcharge genre is not a chemical phenomenon. It is what happens to any line item with a motivated seller and a buyer who cannot check the math, which is to say most of them. The headline changes. The arithmetic does not.

The chemical version of it is a report. Every quarter, a should-cost decomposition of the major product families against the public indices they ride: the model in this piece, kept current, waiting in a drawer for the morning a deck lands on your desk. For anyone who wants it live instead of quarterly, the same model sits on a platform wired straight to the feeds.

Paid subscribers get the quarterly report starting in Q3 2026. the live platform is for people who hate waiting.

Subscribe now

What the Model Knows (and What I Won’t Tell You)

I am now going to be coy about the model’s internals, and the coyness is load-bearing. Part of what the model is worth is that the vendor knows we can see the cost structure without knowing exactly how finely we can see it. Ambiguity is usually the seller’s asset. This is the rare place it works for you.

What I will tell you is the shape. Every product family reduces to a handful of feedstock families: methanol, ethylene derivatives, aromatic solvents, fatty acids, phosphoric acid, glycols. Corrosion inhibitors lean on fatty acids and aromatics. Demulsifiers are ethylene oxide chemistry. Hydrate inhibitors are methanol wearing a jacket. Each family maps to an index you can pull without anyone’s permission: Methanex posts its North American methanol price monthly, CME lists ethylene futures, Platts assesses the aromatics, and crude is a workable proxy for the petroleum-derived solvents. Above the molecules sit four more layers (blending, logistics, SG&A, margin), and the entire model rests on one observation: a headline touches the first layer and merely waves at the others, which move with ordinary inflation no matter what the news is doing.

The 50/50 fact makes it sharper. Half a chemical contract is not chemistry at all; it is service and knowledge, the tech and the lab. A commodity shock touches, at most, the raw-material slice of the chemistry half. A flat surcharge on the whole invoice therefore repriced the field techs’ time because a commodity index moved, which is a sentence nobody would say out loud, which is why it was a percentage on an invoice instead of a sentence.

Vendor: our costs went up.

You: which of them?

Vendor: ...costs.

That exchange, with a model behind it, is most of the job.

What I will not tell you: the weights, the per-family formulas, the full input lists. But you do not have to take the decomposition on my authority. The same Innospec 10-K that told you how the suppliers buy also tells you what they buy: oleochemicals, ethylene, solvents, amines, alcohols, polyacrylamides. The same families. The suppliers already index their inputs. The surcharge proposes that you, alone in the entire chain, pay spot. Someone has to, I suppose.

The Alternative Market That Isn’t

There is a quieter claim inside every surcharge conversation, usually unspoken, occasionally said out loud when things get tense: we have other places to sell this. That is the vendor’s BATNA, and it deserved a look, so we looked.

The look was unkind. The feedstocks in production chemicals are not oilfield molecules; they are everyone else’s molecules that the oilfield borrows. Packaging consumes roughly 45% of the world’s polyethylene; oilfield demand does not register in ethylene end-use breakdowns. Polyester and PET own about 65% of xylene chemistry and nearly 90% of MEG. Methanol belongs to formaldehyde, construction resins, and Chinese methanol-to-olefins plants. Phosphoric acid is 85% fertilizer. You are not the customer in these markets. You are barely a rounding adjustment on the customer. And the actual owners were, at the moment of the surcharge, mostly not buying: global ethylene operating rates are headed toward 79% against a historical 88%, after the industry added 40 million tonnes of capacity from 2020 to 2025 against 27 million tonnes of demand, and Chinese MTO units were running at 60% utilization. The spikes in the vendor’s deck were supply-shock premiums on structurally oversupplied markets. That is the kind of price that corrects, and the kind of alternative demand that is not standing in line for anyone’s diverted totes.

Two honest exceptions, because the model gets to be honest or it gets to be useless. Phosphoric acid is genuinely tight, and tightening for reasons that have nothing to do with any headline: LFP batteries are pulling purified phosphoric acid toward automakers, who may be 24% of demand by 2030. Tall oil fatty acids are structurally tightening because renewable diesel mandates outbid everyone for the feedstock. Both are real. Both predate the conflict. Both belong in contract repricing at renewal, not in an emergency surcharge wearing the headline’s jacket.

You Can’t Negate Alone

The model won the argument. The model could not end the negotiation, because arguments are not the thing negotiations are made of. There was a person on the other side of the table who now had a problem: he had sent a corporate deck demanding 12%, the customer had just proven 7%, and somewhere above him sat a spreadsheet with his name on it and a surcharge-revenue column that was no longer going to fill itself. If you make his refusal humiliating, you win a quarter and poison a decade. This is a repeated game. Same vendors, same basins, same kids’ soccer sidelines, for the next twenty years.

So you give. Not on price. On things that are cheap to you and look like trophies to him. Term is the classic: a contract extension at current pricing converts your price integrity into his retention story, and retention is what his corporate actually tracks. (The extension costs you nothing if you were staying anyway, and, be honest, you were.) The other is a trial of a higher-strategic product line: pilot the technology he is under pressure to commercialize, at cost, in exchange for the performance data and the reference rights his sales deck is starving for. The field team books a win it can report upward that is worth more internally than the surcharge ever was.

And the choice itself is a diagnostic, which is the mechanism design part, and my favorite part. A vendor in a genuine cost emergency takes cash in any form cash arrives. A vendor that cheerfully trades a 12% surcharge for a contract extension and a case study has told you, in the language of revealed preference, what the surcharge actually was. It was margin expansion with a headline on the cover page. You did not extract that confession. You designed a menu, and he ordered it.

Stepping Over Nickels

There is one more move, and it is the one that closed ours: stop arguing about the size of the nickel and show them the dollar they are stepping over.

Run the materiality math from the vendor’s side. A 12% surcharge on a mid-market chemical program is a couple hundred thousand dollars a year. Against the revenue base of a major service company, that is not a number. It is roughly what the company distributes to its shareholders in the time it takes to schedule the meeting about the surcharge. And the operator on the other side of the table is the growth story. Field people already have an idiom for this exact error, pointed the other direction: one executive described cutting frac chemistry to save a few thousand dollars a day, against a completion burning hundreds of thousands a day, as tripping over dollars to pick up pennies. Turn the idiom around and hand it back across the table. In this negotiation, the vendor is the one stepping over dollars.

What the dollar looks like depends on the operator, and every operator has a version of this story; most never think to tell it. I will not walk through every situation here (mapping yours precisely is what the custom market intelligence reports are for), but a few general ones illustrate the point.

1. The non-Permian oil operator with deep inventory. While the Permian fights Tier 1 exhaustion and consolidation eats the customer list one acquisition at a time, an operator with decades of undrilled locations in an unconsolidated basin, adding rigs while the public companies hold maintenance-mode flat, is one of the only organic growth stories left in the vendor’s North American book. (The major suppliers need North American growth stories right now considerably more than they need surcharge revenue.) The supplier embedded there today rides the basin’s growth without ever re-competing for it.

2. The vertically integrated gas producer. Owns its molecules from wellhead toward market, which makes its revenue, and therefore its chemical program, annuity-grade. In a business where production chemicals are prized precisely because they are insulated from rig-count cycles, the steadiest book of business in the building is not the account you reprice with a rider. It is the account you protect from your own corporate.

3. The ABS-backed operator. Securitized PDP, hedged production, a strict cash waterfall in which LOE sits senior to the noteholders. Chemical cost stability there is not a preference; it is approximately a covenant. And every new financing the operator closes is acreage and production the embedded vendor inherits without a single sales call. A surcharge dispute is a strange way to treat a customer whose corporate structure legally obligates it to be boring and pay you forever.

In each case the message to the supplier is the same, and it is not a threat. It is an invitation to do arithmetic: you can have the surcharge, or you can have the option on everything we are about to become. The vendor that prices the relationship like a line item gets treated like one.

Ours did the arithmetic. The surcharge went to zero. The vendor left with a package of things worth considerably more to it than the surcharge ever was, and everyone reported a win, which is the defining feature of a deal that was never really about costs.

And, as I said at the top, “ours” is not singular. The sequence keeps ending at zero because none of it is magic. It is a model, a market report, and a negotiation sequence, applied before the surcharge’s effective date instead of grieved after it.

British Airways’ customers eventually got their verification technology. It was called the Office of Fair Trading. It required a whistleblower to switch it on, and it arrived four years and $550 million late, refunding one third of the surcharge to people who had long since flown home.

You do not need a regulator. You do not need a whistleblower. You need the model before the deck arrives, because the deck is already written; it is sitting in a drafts folder waiting for the next headline (a tariff schedule, a strait, a storm, a strike), and there will always be a next headline. Cheap talk drifts toward whatever the story will bear. Verifiable claims do not drift. They reconcile.

Talk is cheap. Make them prove it.

Share

That’s it. That’s the whole mechanism behind 180-week lead times, tripled backlog, and the most dislocated equipment market in the history of gas compression. One engine, two buyers, and the uncomfortable math of what’s waiting on the other side of the purchase order.

Buyer one is a hyperscaler. The engine unlocks a GPU cluster that earns money by the hour the instant it switches on. TD Cowen estimates these buyers routinely pay $1 million to $2 million per megawatt for behind-the-meter generation, and up to $3 million to $5 million per megawatt for islanded configurations that bypass the grid entirely. The engine is a rounding error on a $10 billion campus.

Buyer two is a compression operator. Not the same engine, technically. The compressor driver and the generator set are different configurations. But they share the same block casting, the same crankshaft forging, the same production line at Lafayette, the same 18-month queue. Caterpillar decides how many of each to build on that line, and the decision follows the money. The compression variant unlocks a contract priced at $25-31 per horsepower per month. One buyer’s rounding error is the other buyer’s entire capital budget.

Same line. Different engine. Wildly different thing waiting behind it.

The auction never measures the engine. It measures what’s behind it. The compression operator can be the most disciplined, rational, well-run buyer on earth and still lose every single time. He isn’t being outsmarted. He’s being out-staked.

Kalibr tracks every compressor unit across the Lower 48. Of new compression sets above 1,000 HP permitted in 2025, 85.2% were Caterpillar. Waukesha placed 11%. Cummins placed less than 1%. Waukesha’s 11% is the rounding error. Cummins’s share is a rounding error on the rounding error.

The wait-and-see instinct that has served compression operators for twenty years (order when you have the contract, manage working capital prudently, don’t get overextended) is now structurally the losing move. Not because the operator got less disciplined. Because the other buyer’s mistake can only ever be warm: over-order and the balance sheet shrugs, you’ve got backup power anyway. The compression operator’s mistake bites from both directions. Order early and you’re bleeding capex against demand that hasn’t shown up yet. Order late and the engine is 180 weeks out, sitting in line behind a data center, and you can’t serve the contract you just won. Prudence stops being a strategy and becomes a way to lose slowly.

The demand numbers are public. Caterpillar’s large reciprocating engine backlog has grown more than 3.5x since January 2024, to a record $62.7 billion across the enterprise. Power generation sales surged 41% year-over-year in Q1 2026. That number is correct. It is also the less interesting half. The auction logic tells you who wins the engine. The supply chain tells you why there’s an auction at all. There are exactly 33 facilities on earth capable of forging a crankshaft for a Caterpillar 3600-series engine. Zero domestic U.S. foundry capacity for large engine castings. Turbocharger superalloys competing for the same inputs as Boeing and Airbus. The binding constraint on Caterpillar’s engine output isn’t Caterpillar’s assembly lines. It’s a network of foundries, forging houses, and turbocharger specialists scattered across Brazil, India, Finland, and Japan.

Here is the part that is not in the earnings deck.

The Production Reality

Caterpillar does not disclose unit-level production figures for its G3500 and G3600 engine families. It never has. So you have to build the number yourself.

Start from the bottom up. Kalibr’s data shows three engine models account for 90% of every Caterpillar engine going into compression above 1,000 HP: the G3516 (40.6%), the G3608 (28.1%), and the G3606 (21.3%). Everything else is rounding. The compression industry’s Caterpillar dependency is actually a three-engine dependency, which means the engine allocation question everybody is trying to answer really comes down to: how many G3516s and how many G3606/8s does the compression channel get?

Now work from the top down. Jefferies Research estimates that Caterpillar currently produces 1,500 to 2,000 large reciprocating engines annually, with the capacity expansion at Lafayette projected to push that figure above 3,200 by the end of the decade. RBC Capital Markets breaks the P&E segment revenue into its components: the 22% year-over-year growth in Q1 2026 was driven by a 14.5% increase in physical sales volumes and a 1.9% price realization premium. That 14.5% volume growth, applied to the Jefferies baseline, implies mid-single-digit unit growth in the trailing twelve months. Caterpillar is making more engines. The compression industry is getting fewer of them.

Our midpoint: 1,650 large reciprocating engines rolled off Caterpillar’s lines in 2024. In 2026, with the initial phases of the Lafayette expansion contributing efficiency gains (new automated crank milling equipment replacing 40-year-old machines, reducing operator requirements from 30 to 8 per station), that number rises to 2,100.

The question is where those 2,100 engines go.

The Allocation Shift

Historically, Caterpillar’s large engines were split roughly evenly between oil and gas applications (compression, well servicing, drilling) and power generation (prime power, backup, distributed). That balance has broken.

Power generation is growing at three times the rate of oil and gas. RBC Capital Markets reports that power generation has become the largest component of the Power & Energy segment, at 32% of total revenue in 2024, surpassing oil and gas at 29%. The trajectory is clear: in Q3 2025 the power gen growth rate was 33%. By Q1 2026, it was 48%. That is not a trend line. That is a reallocation.

The allocation shift hits the two engine families differently. The G3500 series, particularly the G3516H (Caterpillar’s dominant 1.9 MW data center genset), sees its DC allocation rise from 50% to 70% at peak in 2028. The G3600 series crosses from compression-majority to DC-majority in 2026. It never fully tips back. USA Compression’s management confirmed this on their Q1 2026 earnings call: Caterpillar has no major plans to expand manufacturing for the 3600-series engines used in gas compression.

Our unit model estimates the 2x2 split:

Total production rises 27%. Compression availability drops 6%. The allocation shift eats the entire gain.

For the compression channel, the arithmetic is straightforward. If Caterpillar produces 2,100 large reciprocating engines in 2026, and 58% go to power generation, the compression channel gets 885 engines. Total compression demand sits around 930 engines. That’s a 45-engine deficit, roughly 5%, which doesn’t sound catastrophic until you realize that the deficit is concentrated almost entirely in the 3600 series, the 2,000 to 5,000 HP class that is the backbone of midstream compression.

Kalibr’s data shows why that concentration matters. In the 2,000 to 3,000 HP band, the G3608 is 89% of all new compression sets. Above 2,000 HP, Caterpillar’s total share is 91%. The only reciprocating alternative is Waukesha, at 5%. The 45-engine deficit is not spread across a market with substitutes. It is concentrated in the one horsepower band where substitutes do not exist. All the revenue triangulation and capacity estimates above come down to one engine model in one horsepower band. Whether those engines compress gas or generate electricity for a data center is not a Caterpillar decision. It is a price discovery. And the data center will always pay more.

The Three Chokepoints

Caterpillar announced a $725 million expansion of its Lafayette, Indiana Large Engine Center. Wall Street celebrated. The stock moved. Then someone asked where the crankshafts come from.

Here is what $725 million actually buys. The Lafayette footprint is growing by about 400,000 square feet, a 30% increase to the existing 1.7 million square feet. The other 95% of the 125% capacity increase comes from replacing 40-year-old machining equipment with modern automated systems and debottlenecking internal production flow. Bernstein Research walked the facility and reported the details: new automated crank milling equipment that reduces the operator count from 30 to 8 per station, modernized boring equipment for cylinder blocks, upgraded quality control systems.

Caterpillar is not building a new factory. It is modernizing an old one. And the binding constraints on its output are not inside Lafayette. They are upstream, in the foundries and forging houses that supply the raw castings, forgings, and turbocharger components that Lafayette machines and assembles.

Caterpillar management has been explicit: increasing total throughput is “highly dependent on bringing the external supply base along” with internal facility expansions (which is a polite way of saying: we can build engines as fast as Brazil can cast engine blocks, and not one unit faster). They’ve dedicated significant resources to working with Tier 1 and Tier 2 suppliers on synchronized ramp-ups.

The three component classes that define the production ceiling:

1. Large Custom Castings

Every G3600 engine starts as a multi-ton iron casting: the engine block and cylinder heads. These castings require specialized ductile iron grades, massive high-precision vertical machining centers (only a handful of machine tool builders globally manufacture equipment large enough to process 3600-series blocks), and deep metallurgical expertise.

The United States has zero domestic foundry capacity for these parts. Zero. Every cast iron engine block and cylinder head that goes into a Caterpillar 3500 or 3600 series engine is imported, primarily from Brazil, Mexico, and Europe.

Tupy SA, the world’s largest iron foundry (748,000 tons per year across plants in Brazil and Mexico), is the dominant Tier 1 supplier. Bradesco Corretora estimates that 70% of the medium and heavy truck fleet in the United States uses engine blocks manufactured exclusively by Tupy. For large industrial engines, the supply base narrows further: Craftsman Automation in India, Componenta in Finland, Linamar in Canada, Anhui Yingliu in China. Five foundries, four continents, and zero domestic capacity. The entire domestic heavy truck fleet and the entire large engine industrial base, dependent on a single Brazilian foundry most Americans have never heard of.

Craftsman Automation is building out its Kothavadi foundry in India (it acquired Germany’s Fronberg Guss GmbH to scale from 100-kilogram to 3-tonne components). The new capacity comes online in FY2029. Three years from now. And the order book? Linked primarily to data center power generation demand. Not compression.

The competing demand for large engine block casting capacity: data center backup power generators (directly competing for identical castings), industrial gas turbines (consuming complex casting slots through 2028), marine propulsion engines, mining equipment, and locomotive engines. Every one of these industries is spending at record levels.

2. Large Forged Crankshafts

The crankshaft is the most mechanically stressed component in a reciprocating engine. For the 3600 series, these are multi-ton forged steel assemblies produced on hydraulic or mechanical presses exceeding 8,000 tons of force.

The entire global capacity for large hydraulic forging presses capable of producing these components: 33 facilities. If you decided today to build a new one, you would need $1.0 billion to $1.5 billion for three additional press lines, two years of feasibility studies, and a 6-to-7-year qualification process before a single production crankshaft ships. That puts you in 2034. You would be making crankshafts around the time the current supply crisis has already resolved itself. In the United States, Howmet Aerospace has noted that only four presses exceeding 35,000 tons exist in the entire country.

The key suppliers: Happy Forgings in India (commissioning a new 14,000-ton press, one of only a handful globally with this specific capability), Bharat Forge (large open-die presses, recently diversified into casting via its JS Autocast acquisition), Sumitomo Corporation of America (leading North American supplier), and Park-Ohio Industries (specialized finishing and machining).

CIE Automotive India is the only major supplier capable of manufacturing heavy crankshafts through both casting and forging processes simultaneously. Which makes it either impressively efficient or terrifyingly concentrated, depending on whether you are writing a supplier brochure or a risk assessment. A power grid failure, labor strike, or regional scrap steel shortage affecting either CIE or Bharat Forge would immediately halt production of two out of three critical constraint components for the 3500/3600 platforms.

The competing demand: commercial aerospace programs (monopolizing critical press time for landing gear and structural forgings), surging European defense spending, wind turbine components, nuclear power industry, and railway. All of these industries are fighting for the same finite set of heavy forging presses. All of them are in upcycle. Nobody is waiting.

A former Caterpillar marketing professional told an expert network that critical components like crankshafts have become “nearly impossible to source in the spot market, creating a fragile environment where any minor disruption causes compounding delays.”

3. Specialized Turbochargers

Turbochargers for the 3500 and 3600 series engines operate in an environment most machine shops refuse to enter. The hot section (turbine wheel and housing) is exposed to extreme exhaust temperatures and immense rotational speeds, requiring specialized superalloys: Inconel 718, Hastelloy, Waspaloy. These are the same materials that go into jet engine turbine blades. That is not a metaphor. It is the same alloy, the same specification, and in many cases the same foundry.

The turbocharger supply base is dominated by Garrett Motion (which recently launched its “MEG” very-large-frame turbochargers for industrial and marine applications) and BorgWarner (global market leader, now expanding into stationary power applications), with Japan’s IHI Corp and China’s Weifu High-Technology rounding out a supplier list that counts its members without needing a second hand.

Caterpillar’s turbocharger suppliers must bid against Airbus, Boeing, GE, and Siemens Energy for the exact same aerofoil components and superalloy blades. The rapid expansion of AI data centers has triggered what HSBC Global Investment Research calls a “super-cycle” for heavy-duty industrial gas turbines, and these turbines rely on the identical specialized inputs. Every turbocharger that goes into a Cat 3608 is a turbocharger that didn’t go into a GE gas turbine or a Pratt & Whitney overhaul kit.

China and Russia dominate the raw inputs. Russia’s VSMPO, which historically provided 25% to 30% of the aerospace industry’s titanium, has been effectively cut off since the invasion of Ukraine, forcing the Western industrial base onto Japanese sponge producers already running at capacity. Every turbocharger has a geopolitics problem baked into the alloy.

Garrett Motion’s experience is instructive. When the company attempted to source cast turbine wheels from an $80 billion global aluminum caster in China, the supplier “initially lacked the technical know-how to produce the part.” Garrett spent multiple years teaching the supplier-specific alloy recipes and machining techniques. Eventually, Garrett built its own captive casting facility to ensure supply continuity. An $80 billion company couldn’t make the part. That is the turbocharger supply chain.

What the Model Says

We built a model for Caterpillar’s G3500 and G3600 series engine production that starts from the only thing we can actually observe (P&E segment revenue) and works backward through four independent proxies to get to a number Caterpillar won’t give anyone.

The CAPEX Cross-Check

Caterpillar does not disclose unit shipments. But the public compression companies do disclose capital expenditure and HP additions. This proxy validates the model.

These four companies represent about 25% of the U.S. contract compression market by fleet horsepower. Scaling to the full market implies 872 engines of total compression demand. The model’s base case produces 885 compression engines for 2026. That’s a 1.5% delta to the packager proxy. Close enough.

The Raymond James cross-check arrives from the other direction. Their estimate of 26 Bcf/d of incremental natural gas production growth through 2030 implies 15 million additional horsepower, a 25% increase to the installed compression fleet. At 2,000 HP per engine average, that’s 1,500 engines per year. Our model assumes 1,050 engines of annual demand by 2030. If Raymond James is right, the compression deficit is deeper and longer than our base case projects. That is the model’s most conservative assumption.

The public packagers’ ordering behavior tells you who has already priced this in. USA Compression’s management has noted that their future orders are heavily weighted toward the G3600 series (Kalibr’s vendor-attributed permit data shows zero G3600-class units for USAC in 2025, consistent with Caterpillar’s reallocation away from compression). Kodiak has secured Caterpillar engine deliveries through 2028. Archrock has pre-contracted 85% of its 2026 new-build program.

The demand underpinnings are structural.

Production Forecast and Pricing

Read more

There is something underground. It has value today. It will have value tomorrow. The only question is: which value is higher?

If you sell it today and put the money in a savings account, you earn the interest rate. If you leave it in the ground and sell it next year, you earn whatever the market decides that thing appreciates by over the next twelve months. When the interest rate wins, you extract. When the ground wins, you hold.

In 1931, an economist named Harold Hotelling wrote a paper that formalized this logic for exhaustible resources. The paper is 28 pages of differential equations and it boils down to one sentence: the price of a finite resource in the ground should rise at the rate of interest, because if it rose faster, nobody would extract, and if it rose slower, everyone would extract today. It is the kind of insight that seems obvious once someone states it, which is the hallmark of work that is genuinely important.

For 95 years, the American oil industry operated as though Hotelling was wrong. Not because the math was bad, but because the premise felt irrelevant. There was always more rock. The Spraberry led to the Barnett, the Barnett led to the Marcellus, the Marcellus led to the Wolfcamp, and somewhere in there we learned how to drill sideways and frack the hell out of tight formations that a decade earlier would have been written off as source rock. The denominator (drillable Tier 1 locations) felt infinite. So the industry spent a decade perfecting the numerator: returns per well, free cash flow margins, capital discipline, DSU optimization, all the things that make quarterly earnings presentations look crisp and Wall Street analysts say nice things.

On May 20th, Devon Energy paid $357,129 for a single acre of undeveloped New Mexico desert.

And Hotelling’s Rule arrived in the Permian Basin.

I have been writing about this for a year. In January, I called the Devon-Coterra merger what it was: not a growth story, but an exercise in aggregating depletion. In June 2025, I watched EOG, the company that built its entire identity around organic growth, spend $5.6 billion on Encino because the drill bit alone could no longer solve the problem. This is the third article in a series I did not plan to write. The first two were about the diagnosis. This one is about the prognosis.

Subscribe now

The Part Where Devon Buys the Desert

The Bureau of Land Management held a federal onshore oil and gas lease sale on May 20, 2026. The total receipts were $4.01 billion. Devon Energy accounted for $2.63 billion of that. Sixty-six percent.

The company secured 24 parcels covering approximately 16,300 net undeveloped acres in the core of the Delaware Basin, specifically in Lea and Eddy Counties, New Mexico. The acreage adds roughly 400 net drilling locations normalized to two-mile laterals at an average cost of approximately $161,500 per acre and $6.60 million per inventory well. The net revenue interest is 87.5% (the federal royalty is 12.5%, which is actually more favorable than typical state or fee leases in the region). The primary lease term is ten years. The acreage is directly adjacent to Devon’s existing infrastructure, which means the company can plug these locations into its current midstream facilities without building from scratch.

Those are the mechanical details. Here are the ones that matter.

Recent Permian transactions have priced at $80,000 to $130,000 per acre and $3 to $4 million per well. Devon paid $161,500 per acre on average, with one parcel clearing $357,129. That is a 24% to 102% premium to the 2025 average, depending on which comp you use. And Devon was not alone. Matador Resources spent $1.14 billion in the same sale for 5,154 acres at an average of $221,700 per acre, including one bid of $330,002 per acre for a single 1,000-acre tract. Federal Abstract Company paid $221,800 per acre for six parcels. The entire sale priced at a level that would have been unthinkable 18 months ago.

Here is where it gets interesting. Mizuho ran Devon’s new acreage through its ARCHIE model and found that the geology is Tier 1 (some of the most productive rock in the Delaware), but the full-cycle economics are Tier 2 because the acquisition cost lifts the breakeven by approximately 30%, to roughly $49 per barrel WTI. The rock is world-class. The purchase price turns it into something that merely generates very good returns instead of exceptional ones.

(This is the kind of distinction that matters enormously if you are underwriting the asset and not at all if you are writing the press release.)

But here is what Devon’s management is actually solving for. Prior to this acquisition and the Coterra merger, investors had flagged the duration of Devon’s top-tier inventory as a concern. Post-merger, post-BLM sale, Devon’s sub-$40 breakeven Delaware inventory now exceeds ten years. That is not a growth investment. That is a duration investment. Devon did not buy rock to drill faster. Devon bought rock to drill *longer*.

And that, if you have been paying attention, is Hotelling.

Share

The Part Where the Evidence Stacks Up

Devon’s BLM bid is not an outlier. It is the latest and most expensive data point on a trend line that has been building for two years. And if you stack the signals chronologically, the pattern is less of a trend and more of an acceleration.

Start with the mergers. Devon and Coterra combined to create a $53 billion enterprise. SM Energy acquired Civitas. EOG bought Encino for $5.6 billion. In each case, the stated rationale was some version of “synergies” and “scale,” and in each case the actual rationale was inventory duration. You do not merge two depleting portfolios to grow. You merge them to extend the runway.

Then the international moves. Continental Resources, the company Harold Hamm built in the Bakken, acquired a 90% working interest in Argentina’s Vaca Muerta shale in November 2025 and followed it with a 20% non-operated stake in four additional blocks with Pan American Energy in January 2026, spending $378.3 million on acquisitions in Q1 2026 alone. Continental’s CEO Doug Lawler compared Vaca Muerta directly to the Permian: “It could very easily be another Permian in our mind.” He added a line that should make every domestic operator uncomfortable: “The rock doesn’t know what country it’s in.”

EOG, which I wrote about in June 2025 as the bellwether for organic growth’s capitulation, was awarded a 900,000-acre unconventional oil concession in Abu Dhabi’s Al Dhafra region in May 2025 (100% equity and operatorship during the three-year appraisal phase) and established a tight gas sand joint venture in Bahrain. Drilling commenced in Q3 2025. The company that defined “drill it, earn it, repeat” is now drilling it in the Arabian Gulf.

Kimmeridge Capital Management, one of the most prominent US E&P-focused private equity shops, opened a dedicated office in Abu Dhabi in May 2025 and signed an MOU with Mubadala Energy for natural gas and LNG investments. Quantum Energy Partners partnered with Chevron on a joint bid for Lukoil’s non-Russian international assets, valued at approximately $22 billion, spanning production in Kazakhstan, Iraq, Uzbekistan, and Azerbaijan.

And then there is Diamondback. Kaes Van’t Hof, during the Q4 2025 earnings call in February 2026, said it plainly: “Listen, we’re in a depleting business, right? And we think about inventory every day.” Travis Stice had said it even more bluntly in Q2 2024: “We’re running out of Tier 1 inventory.” By Q1 2026, Van’t Hof was writing in the stockholder letter that the light had turned green on growth, because “the operator with the best inventory quality and the lowest cost structure with the longest inventory depth probably has the right to grow organically.”

Probably. An interesting word to put in a stockholder letter. Probably has the right to grow. That is what passes for confidence when the denominator is going to zero.

There is a 19th-century economist named William Stanley Jevons who could have told you this was coming. In 1865, Jevons observed that as steam engines became more efficient, England used *more* coal, not less. Efficiency made the resource cheaper per unit of work, which made more activities profitable, which increased total consumption. The shale revolution followed Jevons’ script perfectly: longer laterals, tighter spacing, optimized completions, better capital efficiency per well, all of which accelerated the rate at which Tier 1 inventory got consumed. The industry optimized its way into scarcity.

But is there a way to measure this structurally, not just anecdotally?

The team at Novi Intelligence published a paper in April titled “From Reserves to Returns” that quantifies it with a precision I am not going to replicate here, and that I’d recommend to any E&P executive or investor who hasn’t read it. The core metric is what they call the Recycle Ratio: operating cash flow per barrel divided by organic proved developed finding and development cost per barrel. It tells you how much cash you generate from your existing production relative to what it costs to replace that production through the drill bit. It is, fundamentally, a replacement metric.

And their key finding is uncomfortable for anyone still running the capital efficiency playbook: the number one driver of variation in Recycle Ratio between operators is not operating cost structure, not capex per well, not price realizations. It is the quality of the underlying rock. The 3-year weighted average Recycle Ratio for the US E&P peer group fell from 213% in the 2022-2024 period to 190% in 2023-2025. All three companies that were acquired and taken off the board in 2025 (HES, CIVI, VTLE) were in the bottom quartile of Recycle Ratio performance.

The market is selecting for rock quality. And rock quality is a function of inventory. Which brings us back to the denominator.

Kalibr Partners

The Part Where the Industry Forks

For ten years, every E&P in America ran essentially the same playbook. Optimize capital efficiency. Return cash to shareholders. Grow modestly, if at all. The playbook worked because the underlying assumption was true: Tier 1 inventory is abundant enough that the only question is how efficiently you can convert it into free cash flow.

Devon’s $2.63 billion BLM bid is the market telling you that assumption has expired.

So what do you do when the essence of your business is running out?

I think the answer is that there is no longer *one* answer. The industry is forking, and the strategies are divergent, not points on a spectrum. Four distinct playbooks are emerging, and understanding which one your company (or your counterparty) is pursuing is, I would argue, the single most important question in E&P right now.

The Landlord Play.

Some operators have accepted that organic growth is over and are restructuring around maximizing cash flow from existing production. I wrote about this in May with the ABS architecture piece: Diversified Energy has built a machine that acquires PDP assets, securitizes them into bankruptcy-remote SPVs, and generates management equity through OPEX discipline rather than CAPEX growth. The structure rewards cost control with a leverage that would make a hedge fund blush. This is early innings. Cumulative upstream ABS issuance has reached roughly $20 billion. Annual volume scaled from $0.5 billion in 2020 to $4.3 billion in 2025. The economics are straightforward: 6.0% to 6.5% ABS coupon versus 8% to 10% high-yield corporate debt, non-recourse, no borrowing base redetermination. For operators who are honest about the denominator, the Landlord Play is not a retreat. It is a different objective function entirely. But it requires a new operational playbook: optimizing OPEX like you have never optimized it before, because every dollar saved flows directly to equity through the DSCR waterfall.

The Pipeline Play.

BKV, EQT, Antero, Caturus, Commonwealth LNG. These companies are vertically integrating from the wellhead to the demand point: LNG export terminals, data center power generation, industrial end-use. The logic is straightforward for gas (you can contract directly with the consumer of your product) and much harder for oil (the refining intermediary fragments the value chain). Antero’s firm transport contracts for LNG, EQT’s play into data center gas supply, BKV’s direct gas-to-power strategy, Commonwealth’s integrated LNG development: these are all different expressions of the same thesis. If you cannot grow the denominator (inventory), grow the numerator (margin per molecule) by capturing value further downstream. More compression per molecule, longer value chain, higher margin.

The Empire Play.

ExxonMobil, Chevron, and to a lesser extent the companies following Continental’s lead into Argentina and the Middle East. These operators are replicating the Permian operational model internationally: applying unconventional drilling and completion techniques to overseas basins where the geology is promising but the technology has not arrived. Lawler’s line, “the rock doesn’t know what country it’s in,” is the thesis. EOG in Abu Dhabi, Continental in Vaca Muerta and Turkey (where early estimates suggest 6 billion barrels of oil and 12 to 20 Tcf of gas in the Diyarbakir Basin alone), Quantum and Chevron bidding $22 billion for Lukoil’s international portfolio. The advantage here is not domestic inventory. It is the ability to deploy capital efficiency techniques, the very skills honed over the last decade, at global scale.

The Clock Play.

Everyone else. Companies that are still optimizing DSU spacing and well costs on a shrinking domestic inventory base. Capital efficiency is the right answer to a different question. If your drilling locations are 15% less productive than last year’s locations (and the Delaware Basin-wide data says they are declining 6% annually), then a 5% improvement in well costs is not solving the problem. It is slowing the rate at which the problem gets worse.

Optimizing the font on a resume when the job no longer exists.

I am not saying capital efficiency is irrelevant. Diamondback, with its $550-per-lateral-foot well costs and sub-$40 breakeven inventory stretching twelve years, has earned the right to run the efficiency playbook because it solved the denominator first. The question for everyone else is whether they are running the same playbook because it is the right strategy, or because it is the only one they know.

Share

What This Means

For operating teams: the first question is not “how do we drill more efficiently?” It is “which strategy are we actually pursuing, and does our capital allocation reflect it?” If you are running the Landlord Play, your compression and service contracts should be structured for OPEX stability, not CAPEX flexibility. If you are running the Pipeline Play, gathering and processing infrastructure is not a cost center; it is a strategic asset. If you are running the Clock Play and have not yet asked your board which of the other three you are transitioning toward, that conversation is overdue.

For service providers (and this includes compression, which is where Kalibr lives): the strategy fork changes your counterparty risk profile fundamentally. Landlord operators are the most predictable counterparties: cost-disciplined, DSCR-driven, stable contract structures. Pipeline operators are your highest-growth counterparties: more infrastructure per molecule, longer value chains, expanding service needs. Clock operators are the most dangerous counterparties, because their business models have not yet adapted to inventory scarcity, which means their capital allocation is the least predictable. Understanding which strategy your counterparty is pursuing is the first step in any negotiation. Everything else is details.

The theory says that the rational response to a depleting finite resource is for its in-ground value to rise until extraction becomes uneconomical or the market finds a substitute. For 95 years, the American oil industry dodged the second half of that prediction by finding new plays. The plays are running out.

Devon’s bid is not irrational. It is not desperation. It is the market pricing what Hotelling described on paper in 1931, repriced in real time, on 16,300 acres of New Mexico desert.

The numerator was never the problem.

The denominator was always the constraint.

Devon just paid $2.63 billion to extend it.

The question for everyone else is simpler. And harder.

What are you going to drill?

Subscribe now

Nominally Hedged goes out to a small list. If you know someone thinking about improving sales / lowering CAPEX and OPEX — or who should be — forward it. That's the whole ask.

Subscribe now

Last week I ended the quarterly update with a question that I did not answer. It was buried in the Bifurcation section, two paragraphs past the part about KGS signing a 10-year contract extension (still the longest primary term ever disclosed in public compression), in a sentence that was doing a lot of work for something that looked like a throwaway.

The sentence was: “KGS booked $1.2 million in credit loss provisions in Q1, up from zero in Q1 2025. At 98% utilization and record margins, who cannot pay?”

I left it there because I did not have the full picture yet. I do now.

Here is what happened when I pulled the same line item across every public compression vendor’s Q1 2026 10-Q filing. Not the earnings call transcript (where nobody mentioned it). Not the investor deck (which is a marketing document with fonts). The 10-Q, which is the one document that an auditor has reviewed and a CFO has signed and a securities lawyer has blessed, and which, because of all of that, contains the things that companies would rather you did not read carefully.

KGS: $1.2 million in new credit loss provisions. Allowance up from $12.6 million to $14.2 million. USAC: allowance up 85% in a single quarter, from $1.5 million to $2.7 million. NGS: CFO acknowledged DSO was “above our expected level” due to “discrete collection and process-related items” (which is a very polite way of saying “some of our customers are slow-paying and we are not entirely sure our back office noticed quickly enough”), allowance up to $390,000 from $317,000. Even Flowco, which is a different business model and probably should not be in this comparison except that I find it interesting, ticked up from $1.2 million to $1.3 million.

Every vendor posted record margins. Every vendor posted record or near-record utilization. Every vendor also quietly increased what it expects never to collect.

There is a wonderful term in behavioral economics for what the 10-Q just gave you. Revealed preference. Companies say a lot of things on earnings calls. (Mickey McKee said pricing power “will continue to 2027 and beyond,” which is a thing you say when you believe it, but also a thing you say when you need your stock to trade above 9x EBITDA.) The provision line is different. The provision line is what the company tells its auditors when no one is performing for the sell-side. It is the honest answer to the question: of all the revenue you booked, how much of it are you actually going to get?

At 98% utilization, the answer got worse.

Read more

There are, broadly speaking, two ways to make money selling oilfield services. You can be better than the other guy. Or you can arrange things so that there is no other guy. For about thirty years, the service sector tried option one and mostly failed. Competing on quality in a commoditized market is a treadmill: you invest in differentiation, your competitor copies it, pricing resets to marginal cost, and everybody goes home with a 6% EBITDA margin and a management team that describes the year as “challenging but encouraging.” The operator, meanwhile, runs the RFP, picks the cheapest bid that meets spec, and moves on to the next line item. This was a comfortable arrangement for everyone except the people selling the services.

What I want to walk through today is how three major cost categories on your AFE independently arrived at option two: make supply scarce. They did it through completely different mechanisms. They did not coordinate. And the cumulative effect on your well economics is not the sum of the three individual price increases. It is the product. That distinction matters and I will come back to it.

Game theorists have a term for what happens when independent players converge on the same strategy without coordinating: dominant strategy equilibrium. It means that constraining supply is the best move for each player regardless of what the other players do. You do not need a cartel when the incentives are this clean. You just need every participant to independently read the same page of the same textbook, which, as it turns out, they all did.

The frac companies chose it deliberately.

Liberty, ProPetro, Patterson: the diesel fleets are being actively retired even when they are still mechanically capable. 1.1 million hydraulic horsepower retired in 2024 and 2025. Zero replacement equipment built. Active frac spreads have dropped from roughly 280 at the 2022 peak to approximately 200 today. That sounds like a demand story until you look at completion volumes, which have not declined proportionally. What declined is the number of providers willing to show up and compete at spot rates.

The mechanism is elegant in a way that deserves a moment of appreciation. You take the top of the market, lock it into long-term contracts on e-fleet and simul-frac, and call it “technology differentiation” (which it partly is). This strands the bottom of the market with too few bidders to sustain spot pricing. The spot pricing collapse then justifies retiring the diesel fleets (why keep a fleet running that earns 8% returns?), which constrains total capacity, which supports the premium on the differentiated equipment, which funds more retirements. Game theorists call this Kreps-Scheinkman: firms commit to capacity levels first, then compete on price given those commitments. The frac companies figured out that the optimal capacity commitment is less. Just, less. Make less of the commodity. Charge more for what remains. This is supposed to be hard to sustain because someone always cheats and adds capacity. Nobody is cheating. The returns on cheating (deploying a diesel fleet at spot rates into a market with 200 active spreads) are worse than the returns on not cheating. The equilibrium holds.

(If this sounds familiar, it should. Liberty is now spending $3.50 on power generation capital for every $1 on frac. KGS is spending roughly $1.70 on power for every $1 on compression. Two different service categories, same arithmetic: the core business has become the harvest asset that funds a power generation buildout. The vendors converged on the same capital allocation without coordinating, which tells you everything about where the returns are.)

The OCTG mills had it done for them, which is even better.

Tariffs do not reduce global tubular supply. They reduce the *substitutable* supply available to a U.S. operator, which from your perspective is the same thing. Section 232 duties, anti-dumping and countervailing duties on Korean and Chinese product, and the cascading effect on trade flows from allied mills have compressed the viable import pool to a fraction of what it was in 2018.

There is a classic game theory model for this called Bain-Dixit entry deterrence, where the incumbent has to do something expensive to keep competitors out: build excess capacity, price aggressively, invest in switching cost barriers. The OCTG mills got a much better deal. The government did the deterring for free. The domestic mills with integrated threading and LML capability now have something close to a regional monopoly on production casing, not because they earned it through competitive excellence, but because the trade policy environment handed it to them. (This is not a criticism. If someone offers you a monopoly, you take it. That is also in the textbook.) One market became two: premium grades requiring specifications, certifications, and institutional relationships became calcified domestic territory. Commodity grades remained contestable but heavily penalized on imported substrate. The strategic question (and the one worth watching) is what the mills do with the rents. The answer is visible in every quote you have received in the last twelve months.

The compression vendors had it done for them, and then, remarkably, chose to make it worse.

Caterpillar’s 3600 in-line engine backlog created the constraint. Cat is reallocating production capacity toward power generation turbines, where a single hyperscaler signs a contract worth more than the entire annual output of the compression aftermarket. The compression companies did not ask for a 180-week lead time. But they recognized its strategic value immediately: when nobody can build a new unit for three and a half years, every installed unit is a monopoly franchise that reprices at renewal.

The textbook response to an upstream supply constraint is to diversify your supply chain. Find alternative inputs. Develop substitutes. Invest in the bottleneck. What the compression vendors actually did was the opposite. KGS spent $675 million buying a distributed power company that competes for the same Caterpillar engines that build its compression units. (Read that sentence again slowly.) USAC acquired J-W Power and redirected its manufacturing capacity toward internal fleet growth rather than third-party sales. Archrock pulled back EMD orders because the grid cannot support them, further concentrating demand on gas-drive engines with the longest lead times. The exogenous constraint arrived, and every vendor chose to deepen it.

In game theory terms, this is a Cournot model with an exogenous capacity ceiling, except the ceiling is set by someone (Caterpillar, responding to Microsoft) who does not know or care what a compression unit costs to rent per horsepower per month. The compression vendors are price-takers on capacity and price-setters on rental rates. That is an extraordinarily comfortable position to find yourself in, and they are behaving accordingly.

Three cost categories. Three mechanisms. One dominant strategy. The operator’s instinct to treat each of these as a separate procurement problem, to run three different RFPs and manage three different vendor relationships as though they operate in independent markets, misses that they are three expressions of the same structural shift. The supply side read the textbook. Every chapter. Including the one about what happens to the buyer.

What happened to the buyer is this: for two decades, the operator was the Stackelberg leader. You moved first (set the budget, defined the program, issued the RFP) and the service stack’s best response was already constrained by your choice. That is over. The service providers and their upstream input markets now set the equilibrium. You observe and respond. The math has not changed. The player who moves first still captures the surplus.

The player who moves first is no longer you.

This is the context in which Q1 2026 compression earnings should be read. Not as a quarterly update on three companies. As the most developed case study of a structural shift that has already reached your frac spread and your casing program, and that will define your F&D trajectory for the next three to five years.

Subscribe now

The Lead Time

The single most important number in the Q1 filings is not a margin or a utilization rate.

Caterpillar 3600 in-line engine lead times extended from 110 to 120 weeks last quarter to 160 to 180 weeks this quarter. An additional full year added in 90 days. A new order placed today arrives in Q4 2029. Solar turbine lead times followed: 102 to 106 weeks in Q4, 146 to 150 weeks in Q1.

This is not a temporary disruption. This is Caterpillar structurally reallocating production capacity toward power generation, where a single customer (Microsoft, Google, Meta) signs a contract worth more than the entire annual output of the compression aftermarket. The compression industry became a price-taker in its own supply chain, and the price is set by a hyperscaler whose cost of downtime is denominated in lost AI training runs rather than deferred production.

(Next week, I am publishing what might be the most fun piece I have written for this newsletter: a complete breakdown of Caterpillar, their supplier network, and a bottoms-up supply side model of where the engines are actually going. The lead time number is the symptom. The model is the diagnosis.)

The quarterly scorecard reflects the downstream consequence.

Archrock

$24.27/HP/month (up 8% YoY). Contract ops margin at 72%, highest clean quarterly record. EBITDA $221 million. Utilization 95%. Leverage 2.6x. CFO Doug Aron retiring by year-end.

Kodiak

$23.31/HP/month (up 3.7% YoY), targeting $24 by year-end. Contract services margin at 70.6%, seventh consecutive quarterly record. EBITDA $190 million. Utilization 98%. 10-year contract extension signed with a top customer (unprecedented in the industry). Zero share repurchases.

USA Compression

$22.73/HP/month (up 8% YoY; 4.7% from J-W accretive mix, 3.2% organic). EBITDA $188.6 million, company record. Utilization 91.9%, diluted by 200,000 HP of J-W idle inventory. Leverage 3.74x, hitting the 3.75x target a quarter early. 2026 new HP now 90%+ contracted, up from 50% at Q4.

Blended

$23.44 average revenue per HP per month. 95% average utilization. 69% average adjusted gross margin. All at or near record levels.

In Q4, I told you to quote Brad Childers’ earnings call language in every negotiation. “Your CEO told investors the industry has caught up with inflation.” We called it the comp. We said to use it like a real estate agent uses a below-asking sale.

That weapon lasted 90 days.

Childers’ Q1 tone shifted. No more “more modest.” Instead: “very happy with overall pricing in the market.” Revenue per HP per month moved higher sequentially and year over year, contradicting the Q4 moderation signal. KGS went further. Mickey McKee stated pricing power “will continue to 2027 and beyond.” Supply chain physics overwrote the earnings call narrative in a single quarter. Cat lead times added a full year. The supply constraint that underpins pricing power got dramatically worse between Q4 and Q1.

The lesson is not “find a new quote.” The lesson is: stop anchoring to vendor rhetoric entirely. Earnings call language is a lagging indicator with a 90-day shelf life. Lead times are the leading indicator. And lead times just told you the supply response is not coming before 2029.

The Bifurcation

There is a number in KGS’s Q1 filing that tells you more about the next two years of compression negotiations than anything management said on the earnings call.

Month-to-month horsepower exposure jumped from 9% of fleet in Q4 to 14% in Q1. A 500 basis point increase in a single quarter. That is 615,000 HP now operating on contractually renewable terms, up from roughly 400,000 HP three months ago.

In the same quarter, KGS signed a 10-year compression services contract extension with a top customer. The longest primary term ever disclosed in the contract compression industry. A second 10-year extension is in process.

These two facts are only reconcilable if the customer base is splitting in two.

Tier 1: The Lock-In.

The top 10 KGS customers (S&P 500 constituents, investment-grade credits, 50%+ of contract services revenue) are voluntarily locking in 7 to 10-year terms at peak-cycle rates. These are not naive counterparties. They have modeled the supply constraint. They have concluded that current rates are the floor, not the ceiling, because the supply response is 4+ years away and the engines that would build new units are going to data centers. Certainty of supply at current rates beats the optionality of going short when the alternative is a 3.5-year wait for replacement equipment.

Tier 2: The Walkaway.

Mid-market operators are rolling off primary terms and not re-upping at the terms offered. They are going month-to-month. That 615,000 HP on MTM at 98% utilization is not idle iron. These are operating units generating revenue. The operators attached to those units are choosing short-term flexibility over long-term commitment. Some cannot commit because their planning horizon does not extend to 2033. Some are balking at the rate. Some are simply declining to sign paperwork.

USAC tells the same story from a different angle. Month-to-month revenue now represents 25.4% of contract operations revenue ($78.5 million in Q1), a 74% increase year over year. The highest MTM exposure in the sector.

And there is a third signal, quieter but worth noting. KGS booked $1.2 million in credit loss provisions in Q1, up from zero in Q1 2025. The allowance for credit losses rose from $12.6 million to $14.2 million. At 98% utilization and record margins, who cannot pay?

Here is why this matters for your next negotiation, and why Tier 2 is the better position to be in right now.

Read more

This is not a criticism. It is a business model. David Einhorn figured this out at Greenlight Capital back in 2018, when his investors wanted to leave and he said (I am paraphrasing) no. The investors complained to the Wall Street Journal. Einhorn kept the money (To be clear - if your investors are complaining to the press about how onerous your liquidity terms are, you made the right call on the liquidity terms).

Private credit scaled this insight to $2.1 trillion. The pitch was elegant: we raise long-term locked-up capital, we lend it long-term to borrowers, and because the money is locked up, we cannot be forced to sell assets at the worst possible moment. We are not a bank. We are not runnable. Sleep well.

Then Blue Owl’s investors stopped sleeping well.

In February 2026, Blue Owl permanently gated redemptions on its $1.6 billion OBDC II retail vehicle after withdrawal requests surged 200%. The firm sold $1.4 billion in loans at 99.7 cents on the dollar to CalPERS, OMERS, BC Investment Management, and (notably) its own insurance arm, Kuvare. Blue Owl’s co-president, Craig Packer, went on CNBC and said: “We’re not halting redemptions, we’re just changing the form, and if anything, we’re accelerating redemptions.”

One way to read that sentence is that Blue Owl found a creative solution to a liquidity problem. Another way to read it is that the structural run-proofing that justified the entire asset class is being tested, and the test is producing sentences like that one.

Blue Owl was not alone. BlackRock TCP Capital reported a 19% NAV decline in Q4 2025. Blackstone’s BCRED faced $3.8 billion in redemptions (7.9% of total assets) in Q1 2026. Moody’s revised its entire BDC sector outlook to negative. Payment-in-kind interest hit 12.8% across BDCs. Median debt-to-EBITDA ratios in private credit climbed from 5.2x in 2020 to 6.5x. Portfolios are estimated to be 85% covenant-lite.

And so $543 billion in undeployed private credit capital needs somewhere to go. Blue Owl’s own management described the destination: asset-based finance, which they called “a $7 trillion market only 5% penetrated by private solutions today.” KKR scaled its ABF book to $74 billion (40% year-over-year growth). Blackstone’s infrastructure and asset-based credit platform hit $107 billion (35% growth). Ares, Apollo, Carlyle: all building permanent platforms to deploy capital into hard-asset-backed structures.

The rotation is not subtle. And one of its most interesting destinations is a corner of the oil and gas capital markets that almost no one outside of structured finance is paying attention to.

Subscribe now

The Thing About Cash Flows

If you have seen *The Big Short* (and if you are reading this newsletter, the probability is high), you know the basic architecture. An asset-backed security is what happens when something throws off cash predictably enough that an investment bank can package those cash flows into a bond, get a ratings agency to bless it, and sell it to an insurance company that needs yield.

The “something” can be almost anything. Mortgages, obviously (that one went well). Music royalties: David Bowie securitized 25 of his pre-1990 albums in 1997 for $55 million at a 7.9% coupon with a AAA rating. Then Napster arrived, the cash flows evaporated, Moody’s downgraded the bonds to Baa3 (one notch above junk), and the music ABS market went dormant for a decade. Then streaming happened, consumption patterns stabilized, and Apollo structured a $1.765 billion securitization for Concord Music in 2025 backed by 1.3 million copyrights (including catalogs from The Beatles to Taylor Swift), achieving an A+ rating from KBRA. The asset class came back from the dead because the underlying cash flows became predictable again.

Aircraft leases. The $4 billion Airplanes Group transaction in 1996 was the largest ABS deal ever completed at the time. By 2025, the market produced 16 transactions totaling $9.5 billion. Solar panels: SolarCity issued the first solar ABS in 2013 ($54.4 million, BBB+, 4.8% coupon). The sector now clears over $2 billion annually. Railcar leases. Trade receivables. Auto loans. Equipment finance.

And now: proved developed producing (PDP) oil and gas wells.

The structure is more elegant than it sounds. An operator transfers PDP assets into a bankruptcy-remote special purpose vehicle via a true sale, which legally isolates the cash flows from the parent company’s credit risk. The SPV issues monthly-pay amortizing notes with weighted-average lives of 7 to 10 years (legal finals out to 2037-2044), sculpted to mirror the natural production decline curve of the underlying wells. The issuer hedges 80% to 95% of expected production for five to eight years, which transforms volatile hydrocarbon revenue into something that looks, to a fixed-income investor, remarkably like a mortgage payment.

The waterfall is strict: taxes and LOE first, then hedge settlements, then senior interest and scheduled principal, then reserve top-ups, then junior tranches, then (and only then) residual equity to the management team. Performance triggers tied to debt service coverage ratios, loan-to-value, and production tracking tests accelerate amortization if things deteriorate. The structure does not trust anyone. That is the point.

Share

Where Oil and Gas Sits

Diversified Energy Company executed the first upstream ABS in November 2019: a $200 million, 10-year amortizing note with a 5% coupon. It was a proof of concept. Six years later, DEC has issued over $3 billion in ABS notes, maintains 85% rolling hedge protection on a five-year basis, operates at a 10% base decline rate with a gas breakeven of $1.80 to $2.00/mcf, and carries a 99% financial compliance track record across its securitized portfolio.

The market followed. Jonah Energy has completed seven issuances generating over $3 billion in total proceeds, including the largest single PDP securitization to date ($750 million, October 2022). Cumulative upstream ABS issuance has reached approximately $20 billion across more than 15 issuers. Annual volume scaled from $0.5 billion in 2020 to $4.3 billion in 2025.

The spread compression tells you where the market is on the maturity curve. DEC’s inaugural deals (2019-2020) priced at approximately 425 basis points over benchmark. The multi-tranche structures from 2021 to 2023 came in around 325 basis points. The broadly marketed variable funding notes from 2024 to 2026 are clearing at roughly 250 basis points. If that trajectory looks familiar, it should. Solar ABS followed almost exactly the same path: early proof-of-concept deals pricing above 300 basis points, compressing to 200-250 basis points as institutional comfort deepened. Late-cycle aircraft ABS trades at approximately T+125 basis points. Oil and gas is early-to-mid-cycle, tracking the solar playbook.

The economics are straightforward. ABS A-tranche coupons price at 6.0% to 6.5%, versus 8% to 10% for high-yield corporate debt. Advance rates run 70% to 75% of PV-10 value, compared to 60% to 65% for traditional reserve-based lending. The debt is non-recourse. There is no semi-annual borrowing base redetermination. No refinancing cliff.

The buyer base has institutionalized rapidly. Insurance and financial services firms now represent 56.9% of allocated capital. Asset managers account for 37.4%. Private credit and alternative funds hold 4.1%. Banks: 1.6%. The insurance bid is the structural anchor, driven by the same thing that drives every insurance allocation decision: long-duration, investment-grade, self-amortizing cash flows that match multi-decade liability profiles. DEC’s most recent issuance (ABS XI) attracted over 30 unique institutional investors. In 2019, the first deal went to a single buyer.

Kalibr Partners

The Management Math

Here is the part that matters if you are an operator, a service provider, or anyone who negotiates with either.

ABS entities are not normal counterparties. The financial structure creates a management team that is economically levered to operating cost discipline in a way that traditional E&P equity holders or PE-backed operators simply are not.

The mechanism: every dollar saved in lease operating expense flows directly into securitized net cash flow, which raises the debt service coverage ratio. Higher DSCR lowers the mandatory cash sweep percentage, which releases trapped cash to the management team as residual equity distributions.

Take a base case: 1,200 MMcfe/month production, $3.50/Mcfe realized price (with hedges), $1.40/Mcfe base OPEX, $2.20 million/month debt service. Securitized net cash flow: $2.42 million. DSCR: 1.10x. At that level, the structure mandates a 100% excess cash sweep. Residual to management: zero.

Now reduce OPEX by 10%. SNCF increases by roughly $0.17 million per month. DSCR rises to 1.18x. The sweep drops from 100% to 50%. Residual to management: $2.3 million per year. And if you can nudge DSCR to 1.25x or above, sweep rates fall to 0% to 25%, and the equity take expands dramatically.

This is something close to equity beta on cost control. Fixed debt service plus variable net cash flow equals a high-leverage outcome for the people running the assets.

DEC’s acquisition of Camino Natural Resources (announced May 2026, $1.18 billion) demonstrates the programmatic model. DEC contributed $210 million from its credit facility. Carlyle took a 60% majority equity interest in the SPV. DEC retained 40%, plus 100% of the undeveloped acreage, plus operating and management fees. The ABS debt is deconsolidated from DEC’s balance sheet. DEC retired $92 million in outstanding ABS principal in Q1 2026 alone. The machine acquires, securitizes, optimizes, builds equity, and repeats. In 2025, DEC signed 57 NDAs, submitted 24 bids, and closed one deal. The pipeline is $125 billion.

Subscribe now

What This Means

For operating teams, the implications are structural, not incremental.

1. Cost underwriting becomes the ballgame. I know from experience on the A&D side that fixed and variable cost analysis typically gets a quick pass using the LOE statements in the virtual data room. Most of the analytical energy goes to PUDs and alpha created at the bit. That approach made sense when the capital structure rewarded production growth. In an ABS structure, where management’s economics are a direct function of DSCR, the cost basis is the single most important variable in the deal model. Get it wrong and the sweep traps all the residual.

   Kalibr recommends a triangulation method: LOE statements from the data room, a physical-asset-based bottoms-up reconstruction of the cost structure (I am biased here, since this is what I have invested to build, but I also think it is correct), and public offset proxy analysis to reduce error. Three independent estimates. Converge on a defensible number. The legal and advisory fees on these deals are substantial, so the deal size needs to justify the aggregate cost. You cannot afford to get the underwriting wrong on a structure where every dollar of OPEX error compounds through the waterfall.

2. Systematic cost reduction is not optional. It is the primary equity lever. The only way management makes money in an ABS structure is driving cost out of the system. Every improvement in LOE (compression optimization, chemical renegotiation, automation, predictive maintenance) reduces the sweep and waterfalls more cash to management. DEC demonstrated this when it integrated Maverick Natural Resources: the team identified and eliminated $150,000 per month in wellhead compression rentals by leveraging existing Diversified infrastructure. That is $1.8 million per year, flowing directly through SNCF into management’s residual.

3. G&A is under a microscope. These deals are scrutinizing overhead in ways traditional A&D never did. The question is: how many wells can I run per engineer while holding to the cost discipline above? This is where AI belongs. Not as a buzzword in an investor deck. In the hands of subject matter expert engineers who can use it to monitor more wells, pattern-match best practices from adjacent industries, and reduce the technical headcount required to operate a given portfolio. Fewer engineers running more wells, armed with better tools. That is the structure this capital stack rewards.

For service providers, the calculus is different but equally clear.

1. Understand the growth vector. Companies embracing ABS will keep doing deals. It is purely a cost-of-capital game: once you have proven the structure works, you acquire more assets into the platform. DEC sees $125 billion in actionable deal flow over the next five years. Identifying these companies early and securing first right of refusal on new assets gives you a revenue stream that is diversified by basin (Appalachia, Oklahoma, Permian, with Bakken, DJ, and Eagle Ford on the horizon) and anchored to a counterparty with structural incentives to maintain the relationship.

2. Understand the incentive structure. These operators do not want your rate sheet. They want solutions that lower their LOE, because every dollar you help them save expands their residual equity through the waterfall. A vendor who helps the management team push DSCR above 1.25x is creating measurable economic value. That is a fundamentally different commercial conversation than negotiating monthly rates. Come with the cost reduction plan. The capital stack will do the rest.

Let's Go Deeper

The capital rotating out of private credit direct lending and into asset-based finance is not a temporary dislocation. Blue Owl, KKR, Blackstone, Ares, Apollo, and Carlyle are building permanent platforms. The banks are leaving. The void is being filled by a capital stack that is structurally different from the one operators and service providers are accustomed to negotiating against.

The counterparty across the table increasingly lives inside an ABS waterfall. Their economics, their incentives, their tolerance for cost, and their appetite for growth are all functions of that structure. If you understand it, you can position for it. If you do not, you are negotiating against a set of incentives you cannot see.

Understand the capital stack. It is the counterparty.

Nominally Hedged goes out to a small list. If you know someone thinking about improving sales / lowering CAPEX and OPEX — or who should be — forward it. That's the whole ask.

Share

There is a concept in corporate finance called the conglomerate discount. A company that owns a collection of unrelated businesses trades at a lower multiple than each business would command independently, because investors don’t trust management to allocate capital as well as the market would. The prescription is simple. Sell the non-core assets. Return the capital. Focus.

For the past decade, the Street has applied conglomerate discount logic to the Denver-Julesburg Basin.

Colorado was the non-core asset. Senate Bill 19-181 rewrote the state’s mandate from fostering energy development to aggressively regulating it. 2,000-foot setback rules. “Severe” ozone nonattainment classification. Permitting timelines of seven to eight months while Texas secured equivalent approvals in three days. Investors applied a “Colorado discount” to every DJ-weighted producer, demanding higher risk premiums and lower multiples. The prescription was the same one every activist gives every conglomerate: sell the DJ. Focus on the Permian. Simplify.

The echo for portfolio rationalization hasn’t stopped. Kimmeridge released a letter to Devon’s board just this week demanding “an accelerated program of non-core asset divestitures.” The playbook is evergreen. The logic is clean.

The conclusion might also be wrong.

The thing about conglomerate discount logic that nobody mentions in the activist presentation: the prescription assumes fungibility. Sell the “bad” asset, redeploy into something “better,” watch the multiple expand. This works precisely as long as “something better” exists. The Permian is running out of Tier 1 rock. Not catastrophically, but in the way that matters to anyone building a five-year drilling program: the next well is slightly worse than the last one. Bernstein called it in early 2026. Eagle Ford growth is unlikely through 2030. The Bakken is price-sensitive and declining. Delaware Basin productivity declined 6% year-over-year broadly, 15% for specific operators.

When the “better” alternative is deteriorating, the discount thesis collapses. The DJ isn’t dragging down your multiple because it’s a bad asset. It’s dragging down your multiple because the market hasn’t recalculated what “better” means in a world where everyone is running out of good rock at roughly the same time.

I’ll be honest: I thought Chevron and Civitas would eventually exit the DJ. Two deals changed my mind.

SM Energy acquired Civitas for $4.6 billion not because SM loved Colorado (SM’s CFO cited “comfort with Colorado” at the Bank of America conference, which is the kind of thing you say when you’re trying to convince yourself as much as the audience). SM bought Civitas because SM’s Midland inventory was severely depleted. Devon acquired Coterra for the same structural reason. Not synergy. Inventory duration. When your Delaware Basin reserve life is 8.1 years against an 11.5-year peer average, you’re not buying efficiency. You’re buying time.

These aren’t synergy deals. They’re admissions that the replacement hypothesis failed. The Street demanded asset rationalization. The geology demanded inventory preservation. The geology won.

And the DJ Basin (the asset everyone was told to sell) produces 40 barrels of water per well compared to 400 in the Permian. Recent vintage wells show slight productivity increases versus the seven-year average, bucking degradation trends across every other major play. 75% of Chevron’s DJ locations break even below $50/barrel. Laterals have lengthened 29% since 2018. The economics aren’t aspirational. They’re proven, improving on the margin, while the Permian’s marginal well gets slightly worse every quarter.

We score this probabilistically. Across five evidence streams, our assessment is that DJ Basin sentiment is structurally strengthening at 70% probability (confidence interval 60-80%). Three of five dimensions trending “improving.” The catalyst is not Colorado getting better. It is competing basins getting worse.

The legislative truce helps. Senate Bills 24-229 and 24-230 paused anti-drilling ballot initiatives through 2028. Permitting times declined 50% in two years. Colorado permit approvals grew at 23% CAGR since 2021. The “Colorado discount” is still priced in. The regulatory risk that justified it is largely gone (for a couple of years, atleast).

This is Kalibr’s first basin-level compression intelligence report. We mapped every active compressor in the DJ: 1,150 units, 476 facilities, ~1.275 million HP. We classified the fleet by vendor. We built a bottom-up demand model showing 177,000 HP of new demand in 2026. What we found is a market where compression demand has structurally decoupled from rig count, and where the four largest vendors are positioned very differently for what comes next.

See The Platform

The Part Where We Count the Compressors

1,150 active units. 476 facilities. ~1.275 million HP. Average engine age: 14.7 years.

Context: KGS’s entire national fleet is 4.5 million HP. AROC runs 3.7 million. USAC, 3.9 million. The DJ represents 5-7% of the US contract compression market, concentrated in a single state with the most stringent emissions regime in the Lower 48. The Permian has more iron. The Haynesville has bigger units. No basin has Colorado’s air quality rules.

The fleet has a split personality. Unit counts are eerily even across HP bands (~25% each), but the 1,500+ HP band carries 52% of all installed horsepower from 24% of units. If you think of the DJ as a “small wellhead compression” market, you’re missing that half the basin’s capacity sits in centralized stations with 5+ compressors per facility (10% of sites, 48% of HP). Caterpillar powers 73.8% of the fleet. Waukesha runs the big iron at 16.1% (averaging 1,581 HP per unit). Cummins handles gas lift at 5.9% (averaging 359 HP).

The fleet is old, and that’s not an accident. When operators view a basin as temporary (the “we’ll exit Colorado eventually” thesis), they redeploy older units and run them until the contracts end. That logic persisted for a decade. Now the contracts aren’t ending. The fleet is still old. 294 units from the 2010-2014 horizontal boom are approaching major overhaul economics. 121 units have already crossed or are approaching the 20-year replacement threshold.

The annual additions data shows a market in transition. 2024 saw 244 units set (68% above average) but average HP collapsed to 858 as gas lift installations surged to 43 units. 2025 corrected to 120 units with average HP recovering to 1,118. The question: was 2024 an anomaly or the front edge of a structural gas lift build cycle as the 2018-2019 vintage wells (1,200-1,400 spuds/year) enter the decline phase where artificial lift becomes economic?

The basin splits 50/50 on HP between midstream-owned and E&P-contracted equipment, but the character is completely different: 313 midstream units averaging 2,105 HP versus 837 E&P units averaging 768 HP. The 287 midstream-owned units (~593,000 HP) sit permanently outside the contract compression market. Which means the market AROC, KGS, and USAC actually compete in is roughly 780 units and 596,000 HP. Smaller than it looks.

Who Wins That Market

The most valuable piece of intelligence in compression is the one nobody has: which vendor operates which unit, at which facility, for which operator. Every basin-level analysis you’ve seen treats the fleet as one undifferentiated pool because that’s as far as the data gets them.

Kalibr’s proprietary classification model resolves this. We’ve built a machine learning system that attributes vendor identity across compression fleets using facility-level spatial patterns, operator relationship signatures, and equipment configuration data. In the DJ, the model classifies 84% of units by vendor. This is the foundation that makes everything else possible: market share, runtime associations, contract term exposure. Without vendor attribution, you have a fleet count. With it, you have a competitive intelligence map.

Subscribe now

Read more

Devon Energy can call any compression vendor in the Permian and get a yes. KGS has 4.5 million horsepower. Archrock has 4.8 million. USAC, post-J-W, has 4.4 million.

Devon chose NGS. NGS has 563,000.

The CEO was specific about what drove the expansion: “as they understood some of the capabilities of our units and some of the data that they would be able to get off of that, that was the primary driver.” That is the management narrative. There is an alternative explanation that fits the data at least as well: when two customers account for 59% of your revenue, you deliver exceptional service to those two customers because the alternative is going out of business. The technology story and the concentration story produce identical observable outcomes (great uptime for Devon). They produce very different predictions about what happens when NGS tries to serve customer number 61 with the same 196 field employees.

That distinction tells you something about this market that the earnings call does not.

We have spent the last three weeks taking apart the Big Three. The conclusion was not subtle: the strongest vendor BATNAs in compression history. If you stopped there, you would conclude that the only game is learning to negotiate within those three relationships. For 87% of the public compression market, that conclusion is correct.

The other 13% is where the leverage lives. And those vendors are, to put it precisely, weird. Not “might go bankrupt” weird. “Each one has a structural vulnerability the Big Three have spent two years eliminating” weird. The interesting kind.

This week: three niche players that do not fit the Big Three template and do not fit each other. A compression manufacturer with a rental fleet it is trying to grow into something the market takes seriously, whose $30/HP/month headline metric collapses the moment you decompose it by fleet composition (Slide 16 of the investor presentation is doing a remarkable amount of work for a single chart). The fastest-growing public compression fleet in the sector, growing EBITDA at 34% CAGR on the strength of a technology narrative the market is valuing at a 24% discount to peers, which is the equity market’s way of saying “we are not sure that is technology.” And a production optimization platform built around HPGL and VRUs that IPO’d 15 months ago, carries four material weaknesses in internal controls (which have already produced financial statement revisions, a detail the investor presentation omits), and just bought an ESP company to get upstream in the artificial lift lifecycle, which is actually the smartest strategic move any vendor in this analysis has made.

None of them are the Big Three. All of them have pressure points the Big Three spent two years eliminating. For operating teams building a BATNA, these are the three phone calls that change every other negotiation you have this year.

What This Means

If you are an operating team, three things:

Read more

Subscribe now

There is a specific problem that only successful companies have, and it does not look like a problem when it is happening. It looks like success, which is why it is so easy to misread.

Suppose you run a business that does one thing extremely well. Every quarter, the business generates a pile of cash. The first few years, the pile disappears back into growth: more units, more customers, more of the thing you are good at. This is healthy. This is what you want. But the core business has a size, and you do not. Eventually the rate at which you generate cash exceeds the rate at which the core business can absorb it, and you are left with a pile of money and nowhere inside the business to put it.

You have three options. Give it back to shareholders, which is honest but concedes that you have become a terminal-value company. Hoard it on the balance sheet, which your shareholders will punish you for. Or find a new market adjacent to the core, where the company’s existing capabilities give it an edge and the next dollar has a better home than a brokerage account. If the CEO has any ambition at all, door number three is the only door. The incentive structure guarantees it.

In 1997, Boeing walked through door number three. Commercial aerospace had a size, Boeing had outgrown it, and McDonnell Douglas offered a defense adjacency that used the same airframes, the same engineering, and the same customer logic. The pitch book was immaculate. The deal was not stupid. It was the textbook answer to the textbook problem of excess capital in a bounded market.

What Boeing did not price into the deal was what happens when a single-answer capital committee becomes a two-answer capital committee. Before the deal, every quarter the committee had one question: what does commercial aerospace need? After the deal, it had two. And the second question was always reasonable, always had a pipeline behind it, always had executives in the room making a perfectly defensible case for why this quarter’s marginal dollar should fund their program. No single decision was wrong. A rebuild on the Renton line gets deferred six months because defense has a tooling deadline. A component interval gets stretched from eighteen thousand hours to twenty-two thousand because the data says you can probably get away with it. Each trade is defensible in the meeting. It is only in the aggregate, over twenty years, that you see what happened, which is that the discipline of the single-answer committee (the one that had produced the fleet that had produced the franchise) had been quietly liquidated one capital allocation decision at a time. The 737 MAX was not a single failure. It was the terminal installment on a bill that had been accruing since 1997.

I want to be precise about what I am saying here, and what I am not. KGS just walked through door number three. The DPS acquisition is a genuinely smart deal in a genuinely attractive market. I admire almost everything Kodiak has built, and I think their AI-driven maintenance program is the most sophisticated thing happening in compression right now. What I am saying is narrower than “this will go badly.” I am saying: the arithmetic of a two-question capital committee creates a specific corridor of risk, and the most likely place that risk materializes is a line item most people are not watching.

Let me show you the math.

Kalibr Partners

The Winning

Here is what winning looks like in contract compression: $1.308 billion in revenue, $715 million in adjusted EBITDA, and a 69.2% Contract Services adjusted gross margin that represents a 247 basis point expansion year-over-year. Fleet utilization at 97.7%, with the large horsepower units (which account for 80% of the fleet by horsepower, because KGS figured out years ago that big iron on long-term contracts is where the economics compound) running north of 99%. Revenue per horsepower per month at $23.10, with management guiding toward $24 by year-end 2026. Free cash flow of $230 million, up 87.7% from 2024. Discretionary cash flow of $462 million, up 23.7%.

The EQT AB private equity sponsor fully exited its 76% stake during 2025. That is the kind of detail people skim over, but it matters: the PE overhang that once gave E&P negotiators a talking point about “sponsor pressure” is gone. The leverage ratio hit the long-stated 3.5x target. The company issued compression’s first 10-year bond at 6.75% and left $1.5 billion of ABL capacity undrawn. The balance sheet has never been cleaner.

And the AI-driven maintenance program, the one I am going to spend considerable time discussing in the context of risk, deserves credit. KGS has shifted from time-based 90-day service intervals to condition-based intervals of 120 to 150 days using a Fleet Reliability Center that monitors units in real-time and custom LLMs that provide field support. This is not marketing language. They are running fewer touches per unit per year, maintaining higher utilization, and doing it at a lower maintenance cost per horsepower than any peer in the sector. We will get to what that means in a moment. First, the report card.

I pulled the data and calculated the same metrics the same way. No adjusted-adjusted numbers, no investor deck cherry-picking. The accountants’ math.

The number that jumps out is not the one you expect. KGS’s 6.4% ROIC trails Archrock’s 10.4% by a wide margin, but that is largely a function of leverage (debt is a denominator, and KGS has more of it). The Contract Services margin is best-in-class, and fleet utilization leads the sector. Those are genuinely impressive operating results.

The number that matters for this article is the maintenance CAPEX row. KGS spends $17.05 per horsepower annually. Archrock spends $23.12. That is a 36% gap. On similar-sized fleets serving similar customer bases in similar basins. Archrock’s approach is the traditional one: spend more, overhaul more frequently, maintain buffer. KGS’s approach is the technological one: spend less, use AI to extend intervals, trust the data. One of these approaches is right and the other is wrong, or (more likely) both are defensible within their respective capital structures, and the question is what happens when the capital structure that supports the lower-spend approach gets stretched by a second mandate.

Share

The Door

On February 5, 2026, KGS announced the acquisition of Distributed Power Solutions for $675 million: $575 million in cash drawn from the ABL facility and $100 million in newly issued stock. DPS brings a 384-megawatt fleet of reciprocating engines and turbines, exclusively powered by Caterpillar, serving data centers, midstream operators, and industrial microgrid customers.

I want to say this clearly: this deal makes sense. The Caterpillar 3516J engine that powers KGS’s compression fleet shares mechanical DNA with the 3516H generator set engine. The 700-plus Cat-certified technicians already know the platform. The spare parts inventory overlaps. The customer adjacency (Permian operators who need behind-the-meter power because the grid will not show up for seven years) is real. RBC raised its price target from $45 to $64. The market cap expanded by $420 million. Northland Securities noted the market is capitalizing the acquired EBITDA at a premium 12x multiple, which is the market’s way of saying: we like this.

The deal was executed at 7.4x 2026 estimated adjusted EBITDA, implying roughly $91 million in near-term earnings. Of the acquired 384 megawatts, 250 are dedicated to data centers, including a 100 MW Virginia contract extending 15 years. Management’s demand pipeline is weighted 9-to-1 in favor of data centers over traditional oil and gas. In a world where grid interconnection queues stretch five-plus years and hyperscalers will pay almost anything for reliable behind-the-meter power, this is a market where the demand curve is doing the selling for you.

But (and there is always a but, this is a newsletter about compression, not a pitch book) there are specifics worth noting. The DPS fleet is concentrated around 16.5 MW Caterpillar SMT130 units. It lacks the larger 35 and 38 MW turbines that entrenched competitors use for utility-scale data center work. Much of the industrial and microgrid portfolio sits on one-to-two-year contracts requiring immediate repricing and extension upon closing. Piper Sandler flagged this directly: outside of the primary 100 MW data center agreement, the portfolio needs urgent commercial work on day one. And pro forma leverage, per Seaport Research Partners, spiked to roughly 4.0x.

This is a fleet that needs capital on arrival. The question is where the capital comes from, which requires meeting the neighbors.

Leave a comment

The Neighbors

Here is something I want people to understand about the distributed power generation market, because the analyst coverage treats it as a homogenous growth opportunity and it is not. The companies already operating at scale in this space are not middling operators collecting a tailwind. They are extremely good at what they do, with structural advantages that took years to build and cannot be replicated by writing a check.

Liberty Energy is the one I keep coming back to. Liberty’s power business (Liberty Power Innovations, through the LAET subsidiary) does something that no other oilfield power entrant does: it manufactures its own equipment. The Forte modular scalable generation systems and Tempo power quality systems are built in-house. When Caterpillar lead times for reciprocating gas gensets stretch past 100 weeks and OEMs are enforcing strict customer selection on allocation slots, Liberty is not standing in the queue. It is running its own factory.

That manufacturing integration matters more than most people appreciate. Liberty has 1.3 gigawatts of power agreements in place, a 400 MW reservation with Vantage Data Centers on 10-to-15-year energy service agreements, and a 3 GW capacity target by 2029 supported by an $813 million 2026 CAPEX budget. Piper Sandler notes that up to 70% of that budget can be funded through project-specific financing, which is a polite way of saying Liberty can build power gen capacity without putting the core frac balance sheet at risk. And the balance sheet: Total Debt to Total Stockholders Equity between 0.14x and 0.30x. I want to make sure that registers. Kodiak is running at 3.5x net debt to EBITDA. Liberty’s leverage ratio has a decimal point in front of it.

Liberty’s 14-year average Cash Return on Capital Invested runs 23%. In 2025 it compressed to 13% as completions activity softened (commodities do that), but 13% on a 0.30x levered balance sheet is a company with the margin of safety to absorb a bad quarter, a pricing war, or a new competitor who wants to buy their way into the neighborhood. The ROIC-to-WACC spread (the gap between what the business earns and what the capital costs) is wide enough to fund mistakes. KGS, at 6.4% ROIC on a 3.5x levered capital base, does not have that same width.

The rest of the power gen field is not empty either. Solaris is targeting 2.2 GW by early 2028 with $300,000 in annualized EBITDA per megawatt. ProPetro’s PROPWR division has a 60 MW hyperscaler contract and hybrid gas-battery systems. Atlas signed a 1.6 GW Global Framework Agreement with Caterpillar at high-teens IRR targets. These are not pilot programs. They are funded businesses with contracted cash flows.

None of these companies are better than KGS at running compression. I am not saying that. What I am saying is that KGS’s compression excellence was built over a decade of single-answer capital allocation, where every dollar was a compression dollar. These power gen competitors have been building their own version of that single-answer discipline in their own market. KGS enters as the new entrant against incumbents who have the fleet scale, the contract relationships, and (in Liberty’s case) the manufacturing capability that KGS will need to spend capital to match.

The risk is not that KGS will be bad at power generation. It is that being competitive might cost more than the capital structure can comfortably absorb. And when you are trying to close a gap against operators who have a 500 basis point CROCI advantage and a leverage ratio with a decimal point in front of it, each incremental dollar of investment feels more urgent than the last.

Subscribe now

The Rigid Triangle

I have been circling the arithmetic. Let me land the plane.

KGS operates under three financial commitments that management has publicly treated as non-negotiable. I believe them on all three, which is precisely the problem.

1. The leverage ceiling. 3.5x consolidated net leverage or below. The DPS acquisition pushes pro forma leverage to approximately 4.0x, with Goldman projecting a return to 3.5x by year-end 2026. This is not guidance you can quietly revise. The 3.5x target underpins the midstream-like valuation multiple the market assigns KGS, the pricing on $2.6 billion in outstanding debt (including compression’s first 10-year bond), and $1.5 billion of undrawn ABL capacity. Breach it and you are not just missing a target. You are repricing the equity.

2. The dividend. $0.49 per share quarterly, $1.96 annualized, recently increased 20%. Approximately $170 million per year. Dividend coverage at 2.6x (versus Archrock’s 4.9x, which is the financial equivalent of never having to worry about your rent). Cutting the dividend in this sector is like a restaurant failing a health inspection: technically survivable, but the customers remember.

3. The growth mandate. 150,000 new large horsepower per year in compression, 750,000 total through 2030. Growth CAPEX guided at $235 to $265 million for 2026 compression alone, before DPS power gen spending. The 2026 new-build order book is 100% pre-contracted. Engine deliveries are being secured for 2027 and 2028 at 100-plus-week lead times. You do not call Caterpillar and cancel because your Q2 EBITDA came in light. This capital is committed.

Three immovable objects. One budget. If growth costs more than planned (because power gen scaling requires incremental capital, or because Cat raises prices again, or because the macro softens and EBITDA undershoots guidance), the money has to come from somewhere.

You cannot borrow it (leverage ceiling). You cannot redirect shareholder returns (dividend). You cannot slow the new-build pipeline (committed).

Maintenance CAPEX, guided at $75 to $85 million for 2026, is the line item with give. It is, if you want to be clinical about it, the only remaining degree of freedom in the capital allocation framework.

Maybe it does not need to give. Maybe KGS’s technology-optimized maintenance approach is genuinely sustainable at $17.05 per horsepower while Archrock’s $23.12 represents structural over-spending. The AI is real. The efficiency gains are documented. KGS management has earned the right to that benefit of the doubt.

But the margin between “everything works” and “something has to give” is narrower than I would like. So I built a model.

Kalibr Partners

The Model

Three scenarios. All three hold the dividend at $170 million and the leverage ceiling at 3.5x. Fleet: 4.62 million horsepower.

1. Base Case. Guidance met. EBITDA: $765 million. Growth and other CAPEX: $300 million. Implied maintenance CAPEX: $80 million ($17.31 per horsepower). This is within baseline. Overhauls on schedule. AI extends intervals without degradation. Everything works. I think this is the most likely outcome.

2. Stress Case. KGS leans into the data center opportunity (management described the behind-the-meter market as “vast” and expressed intent to deploy “as much capital as is reasonable”). Growth CAPEX increases $100 million for power gen. Leverage ceiling prevents ABL borrowing. Dividend is sacred.

   Maintenance CAPEX absorbs the shortfall: $35 million, $7.57 per horsepower. A 53.6% deviation from baseline. At this intensity, major engine overhauls get deferred. Spare parts pools deplete. The Fleet Reliability Center keeps monitoring, but the budget to act on its findings is not there. The technology still works. The budget does not.

3. Bear Case. Macro softens. Archrock’s CEO has already noted that repricing increases going forward will be “more modest,” which is the compression sector’s version of saying the pricing cycle has peaked. EBITDA drops to $680 million. Growth CAPEX, locked by 100-week-lead-time commitments, cannot flex. Maintenance CAPEX is forced to $18 million, $3.89 per horsepower.

   At $3.89 per horsepower, you are not optimizing maintenance schedules with artificial intelligence. You are hoping nothing breaks.

Now here is the layer that connects the competitive landscape to the budget. If KGS’s power gen ROIC comes in at 10% (versus Liberty’s 13% incumbent benchmark, which is itself a down year), the EBITDA gap on the initial $675 million investment is approximately $23.5 million. Closing that gap at degraded unit economics requires $235.8 million in incremental capital.

That is 51% of KGS’s annual discretionary cash flow. On a balance sheet that is already at 3.5x leverage. With a dividend that cannot be cut.

I do not think this is the most likely outcome. But I note that the distance between “base case” and “we need more capital” is uncomfortably short, and the consequences of landing in that corridor are disproportionate to the probability.

Share Kalibr Compression

The Part Where the Contract Becomes the Story

I spend a lot of time reading compression contracts (someone has to, and based on the redlines I see, it is not always the people signing them). The availability guarantee is the clause that matters here, and it is worth understanding precisely because the mechanism is a sequence, not an event.

Every KGS contract contains a mechanical availability guarantee: 95.0% to 98.0%, calculated against all downtime including mechanical shutdowns, routine repairs, and scheduled overhauls. Three consecutive months below threshold and the customer can terminate. That is not ambiguous language buried in a rider. It is the core commercial term.

Deferred maintenance does not produce a dramatic failure. It produces a slow erosion that nobody notices until everyone notices.

First, spare parts pools thin out. The localized inventory that enables a three-to-five-hour field cylinder head swap starts getting allocated instead of stocked. Second, the Fleet Reliability Center keeps flagging potential issues (the AI works regardless of the budget), but when the money for parts and labor is not there to act on the flags, response shifts from proactive to reactive. Third, minor issues that were same-day fixes stretch into multi-day outages because the part is not on location and OEM components have their own lead-time queue. Fourth, unplanned downtime hours compound. The 98% number softens.

Industrial software specialists note that even the best predictive dashboards miss roughly 5% of sporadic failures. When budgets are healthy, that 5% gets absorbed by spare inventory and rapid field response. When budgets tighten, the 5% becomes the leading edge of a reliability decline. And KGS has already pushed intervals from 90 to 120-150 days. The buffer is thinner than it used to be.

The revenue exposure: approximately 9.0% of KGS’s fleet sits on month-to-month contracts (30-to-90-day termination). Management indicates 25% to 30% comes up for renewal annually. Against the $1.26 billion 2026 Contract Services revenue midpoint, roughly $460 million is exposed over the next 12 months.

$460 million in revenue governed by availability guarantees that are directly connected to a maintenance budget that is the designated relief valve for a two-front capital strategy. Maybe the technology holds. But if I had $50 million in annual KGS spend and a contract rolling in 18 months, I would want to understand this math.

Subscribe now

How the Peers Solve This

The question is not whether dual-growth strategies work. Baker Hughes pivoted into power and data center infrastructure and the market rewarded it with a premium multiple. SLB is building a billion-dollar data center solutions business. Companies that can fund adjacencies without straining the core tend to be rewarded handsomely.

The question is whether KGS’s specific capital structure has the flexibility to fund both mandates without one cannibalizing the other.

• Archrock stayed home. No power generation entry (explicitly declined; assets did not meet return and quality criteria). The result: 2.69x leverage with 0.8 turns of headroom below target, $110.7 million in maintenance CAPEX ($23.12 per horsepower, highest in sector), 4.9x dividend coverage, 6.0% senior notes (lowest coupon in compression history), no maturities until 2032. Archrock’s capital allocation answer is: grow the core, protect the maintenance budget, let the balance sheet absorb shocks. It caps the upside. It also structurally bounds the downside.

• Liberty walked through door number three with a fundamentally different balance sheet. 0.14x to 0.30x debt-to-equity. When Liberty needs power gen capital, it decelerates frac growth CAPEX, slows buybacks, and uses project-specific financing. RBC notes explicitly: Liberty has outlined a clear funding path that preserves the estimated $200 million annual frac maintenance budget. Liberty never touches maintenance. Its release valves are everything else.

The difference is not about management quality. Mickey McKee at KGS is as good as anyone running a compression business. The difference is structural. Archrock and Liberty have release valves that are not their core fleet. KGS, at its maximum stated leverage with a recently increased dividend, has fewer options. The capital structure funnels any variance from the base case toward the one budget with flex.

And that budget is the one that produces the uptime that produces the revenue that produces the cash flow that funds the dividend that cannot be cut.

Share

What to Watch

Five metrics. All leading indicators of whether this thesis is playing out.

1. Maintenance CAPEX per horsepower. Baseline: $15.00 to $17.50. Below $15.00 signals departure from predictive schedules. Context: Archrock runs $23.12 per horsepower and generates a 10.4% ROIC. KGS runs $17.05 and generates 6.4%. The AI narrative has to keep working.

2. Capital classification shifts. KGS split growth CAPEX into “growth” and “other” in Q1 2025. If maintenance activities migrate into “other” while the headline maintenance line stays flat, the accounting is doing the work that the spending is not.

3. Operating expense per horsepower. Management credits AI for lowering engine repair costs and lube oil consumption. If per-horsepower OPEX rises sequentially, deferred maintenance is surfacing as reactive field cost. Technology delays maintenance. It does not eliminate it.

4. Large HP utilization language. KGS cites 99%+ for large units. If the descriptor shifts from “operating” to “contracted” or “revenue-generating” utilization, mechanical downtime is entering the fleet. Words are leading indicators.

5. The AI efficiency narrative. If management’s emphasis on AI-driven maintenance intensifies precisely as per-horsepower budgets flatten or decline, ask yourself: is the technology supporting the budget, or is the narrative substituting for it?

What This Means for Operating Teams

I want to end where I started. KGS has built something exceptional. The utilization numbers are real. The Cat relationship is the deepest in compression. The AI maintenance program is genuine competitive advantage, and it may be the thing that makes this entire article a theoretical exercise. I would not bet against this management team.

But I would not bet my contract terms on a capital structure, either. The rigid triangle (leverage, dividend, growth) creates a corridor where the maintenance budget absorbs variance. That corridor may never be entered. If it is, the consequences compound through a specific mechanism: deferred overhauls, thinning spare parts, lengthening response times, softening availability, and eventually, customer leverage during $460 million in near-term renewals.

Operating teams with KGS contracts in the renewal window should be doing three things. First, structuring escalating performance credits tied to trailing three-month mechanical availability (not just the headline guarantee, the running average that triggers termination rights). Second, building month-to-month conversion clauses that activate if trailing availability drops, preserving optionality without forcing premature vendor changes. Third, documenting insourcing economics (replacement at current lead times: approximately $1,250 per horsepower, 7-year payback on new, shorter on opportunistic used 3516s) and surfacing the analysis in renewal discussions. Not as theater. As a BATNA.

This is the analysis that moves from interesting to actionable when you map it against your specific contracts, basins, and renewal timelines. The Leverage Matrix was built for exactly that purpose.

Get a Plan

The 737 MAX was not a design flaw. It was what happens when a two-question capital committee defers one small decision per quarter for twenty years.

KGS is not Boeing. The deal is smart, the team is excellent, and the AI-driven maintenance program may genuinely solve the problem I have spent 4,000 words describing. I hope it does. Permian compression is better because KGS is in it.

But the arithmetic of a two-question capital committee is the same arithmetic regardless of who is running it. And when the margin between “everything works” and “something gives” narrows to the width of a single line item on a quarterly filing, someone should be watching that line item.

Watch the line item. Not the headline.

Nominally Hedged goes out to a small list. If you know someone running compression negotiations this cycle — or who should be — forward it. That's the whole ask.

Share

In real estate, when one house sells below asking in a hot market, every listing on the block gets repriced overnight. The sale itself might be explainable (divorce, relocation, bad roof). Doesn’t matter. The comparable is in the system. Every buyer’s agent in the zip code pulls it up on their iPad at the next open house and says, “Well, the comp says...”

This is obviously compression I’m talking about.

On February 25, Archrock CEO Brad Childers told investors that pricing increases going forward would be “more modest” because the industry has “caught up with inflation.” He meant this as a positive. We are operating at peak profitability. Things are great. The machine is running.

What he actually did was create a comp. The largest compression fleet in the United States (4.79 million horsepower, 70-year operating history, every major basin) just told the public record that the repricing cycle is decelerating. That’s not Archrock’s problem. That’s everyone’s problem. When your KGS rep tells you rates are going up 5% next year, you pull up the Archrock transcript on your iPad and say, “Well, the comp says...”

Last week we took Kodiak apart. Peak BATNA, funded walk-away, systematically eliminated vulnerabilities. KGS is the vendor that does not need your contract.

This week: Archrock and USA Compression. Next week: The rest of the publics (Enerflex, FlowCo, NGS). The picture is more interesting than Kodiak, because these two have cracks Kodiak does not.

Archrock is the national incumbent. It is also the vendor whose CEO just handed every operator in America a negotiating weapon they did not have three months ago. USA Compression just doubled its fleet through acquisition and now faces the largest recontracting wave in its history while simultaneously integrating 594 new employees, a new ERP system, and 300+ new customer relationships it has never managed before.

Two vendors. Two different structures. Two very different pressure points.

What This Means (The Short Version)

If you are an operating team negotiating compression contracts, three things:

Read more

Large horsepower lead times doubled in six months. Roughly 55 weeks at mid-year 2025. 110 to 120 weeks today. The proximate cause is not a temporary supply chain disruption. It is a structural reallocation of industrial manufacturing capacity toward power generation, where demand visibility extends a decade, contract terms run 10 to 15 years, and payment economics are simply more attractive than anything the oilfield can offer.

Kodiak spent $675 million to buy a distributed power company. Permian gas processing plants, unable to access the electrical grid for seven to eight years, are converting what would have been 75,000 horsepower of electric motor-driven compression into gas-engine-driven units, consuming the same engine slots the oilfield needs. The compression industry quietly became a price-taker in its own supply chain.

Meanwhile, the Big Three reported Q4 results that belong in a private equity pitch deck.

Subscribe now

Blended fleet utilization at 95 to 98%. Revenue per horsepower at all-time highs: $23.10 at KGS, $21.69 at USAC, an estimated $23.50 at Archrock. And $2 billion in acquisition capital deployed in six months: KGS absorbing Distributed Power Solutions, USAC folding in J-W Power, Archrock integrating NGCS, and Service Compression acquiring AXIP Energy’s Chapter 11 estate. The competitive landscape has meaningfully thinned in a market where new equipment cannot physically arrive for two years.

What this means for your next renewal is straightforward and uncomfortable: your vendor’s best alternative to your deal has rarely been stronger. At 95%+ utilization with no credible supply response before 2028, the threat to walk away is not a negotiating posture. It is free. Every public vendor deleveraged this year. Archrock dropped from 3.3x to 2.7x and issued the first 10-year bond in compression history at 6%, then used it to push debt maturities past 2032. Enerflex hit 1.0x net leverage on record free cash flow. KGS delivered on its IPO commitment of 3.5x and bought back over $100 million in stock. When your counterparty is simultaneously improving margins, retiring debt, raising dividends, and repurchasing shares, the credibility of their walk-away is not hypothetical. It is funded.

The Signal the Earnings Calls Are Broadcasting

Here is the part the prepared remarks do not want you to hear: vendors are pushing aggressively for longer contract terms, and that behavior is deeply informative.

Kodiak is in active discussions with customers about seven- and ten-year renewals. Terms that would have been unthinkable three years ago. Archrock’s fleet now averages 73 months on location, up 61% since 2021, with large horsepower units averaging eight years. Eighty-five percent of Archrock’s 2026 new-build capacity is pre-contracted. KGS’s order book is full through 2026 and they are booking into 2027.

If vendors believed today’s conditions would persist indefinitely, they would be indifferent to contract length. The same terms would be available at renewal. By expending negotiating capital to lock in current positioning, they are pricing in deterioration they can see but you cannot.

This is the quarterback problem. When the party with peak current value demands a long-term commitment, the rational counterparty goes short.

The structural demand runway is real. Raymond James estimates 26 Bcf/d of incremental U.S. gas growth through 2030, implying 15 million additional compression horsepower, a 25% increase in the installed base. Demand will be enormous. But demand is not the same as vendor leverage. The supply response will come, the utilization plateau will break, and the operators who preserved renegotiation optionality through this constrained window will capture disproportionate value on the other side.

Key Performance Indicators | Q4 2025

Read more

I have officially launched the paid tier of this Substack, with the first deep-dive dropping this Friday. The content pipeline will stay fluid. I would rather give you the best and latest from our proprietary dataset than stick to a rigid calendar. The rough weekly cadence will look like this:

1. Quarterly breakdowns of the market and every service provider in the space

2. Basin-level reports: HP setting trends, potential primary term expiry pools, performance benchmarking by vendor

3. Relative positioning of CAT and other OEM suppliers feeding the ecosystem

4. General trend analysis on compression-adjacent activity

These will refresh every quarter. My goal is to arm you with the data to close more deals and lower LOE. I have always believed that great commercial negotiation is not about one side beating the other on book cost. It is about creating better vendor-operator pairings based on their relative positioning and strategy that, as a byproduct, produce better cost outcomes for both sides. If there is something you want me to dig into, just e-mail me. I am here to serve my readers.

Subscribe now

Easter marks the launch into spring sports in my household, which means I am currently operating as an unpaid Uber driver shuttling between batting practice and field hockey, curating walk-out music for a 7U baseball team whose tastes span Hakuna Matata to Sicko Mode, and trying to maintain the thin membrane of sanity that separates a functioning adult from someone who has just explained to a group of 7-year-olds why we cannot all play shortstop.

It also means explaining to a 7-year-old boy why we cannot do baseball practice, ride our bike, *and* watch a movie all in one night.

This is, whether my son appreciates it or not, game theory. Not the John Nash, Beautiful Mind, write-equations-on-windows version. The real kind. The kind that governs every resource allocation decision from a 7-year-old’s Tuesday evening to a Fortune 500 capital plan.

The principle is simple enough for a child to understand and difficult enough that most companies get it wrong: every time you say yes to something, you are saying no to everything else. Not just the thing you considered and rejected. *Everything* else. The entire universe of alternatives you did not evaluate, did not model, did not even see. All of them are now off the table. And if you say yes to everything? Well, you end up like this. And here is the part that makes it interesting: those alternatives do not vanish. They become available to everyone around you. Your no becomes someone else’s yes. Their yes reshapes a market you may not be watching.

A professor I had in business school used to say: “Tell me how often you say no, and I will tell you how strategic you are being.” I have thought about that line roughly once a week for a decade. It sounds like something you’d embroider on a throw pillow, but it is actually a precise statement about competitive positioning. Strategy is not the things you choose to do. Anyone can make a list of initiatives. Strategy is the things you choose *not* to do, and whether you understood, at the moment you declined them, the full map of what you were giving away.

The hard part is not the choosing. The hard part is tracing the consequences. Because trade-offs do not stay local. They cascade. Your product exit becomes an operator’s procurement constraint. Their constraint becomes a competitor’s market opportunity. That opportunity reshapes demand in an adjacent service category that nobody in the original decision room was even thinking about.

Which is why certain rumors percolating around ESPs could have implications for compression demand.

Let me explain, starting with the aircraft manufacturer that has generated more MBA case studies than a compression vendor has reasons they cannot lower your rate.

Share

Boeing’s Hundred-Billion-Dollar Gift to Airbus

Boeing stopped producing the 757 in late 2004. This was, on paper, a straightforward product rationalization. Sales had slowed. The 737 and 787 programs were where the growth capital was going. The 757 was a tweener: too big for short-haul economics, too small for long-haul prestige. Management made the trade-off.

Nobody panicked. The planes still flew. Airlines had years of useful life left in their existing fleets. The consequences would arrive later.

The 757 fleet aged. Meanwhile, the industry shifted toward flexible point-to-point network strategies, exactly the routes where a 757-class aircraft earned its keep. Boeing had read the market for the 757 as declining. What was actually declining was the hub-and-spoke model the 757 was built for. The aircraft category itself was about to matter more than ever.

Airlines that needed to replace their aging 757s did what airlines do: they bought the best available product for the route economics. They did not care whose logo was on the fuselage. Airbus, which had stayed in the category, was happy to oblige. The A321neo, the A321LR, the A321XLR. Airbus built an entire product family to fill the vacuum that Boeing had voluntarily created.

The mechanism is clean enough for a physics textbook. Manufacturer exits product. Operators run aging fleet. Fleet ages out. Operators shop for replacement. The only credible replacement is a competitor’s product. Market share shifts permanently. What looked like a production decision was actually a gift to Airbus that compounded for two decades. The financial magnitude, measured in lost orders, ceded market share, and strategic positioning surrendered, is comfortably north of a hundred billion dollars.

I thought about the 757 when I heard whispers this week about SLB.

Subscribe now

The RC(1000) and Me

Specifically, I heard rumblings that SLB is discontinuing production of certain RC series pumps as well as other legacy gas handling equipment.

That one caught my attention. Let me tell you why.

Back in my days as an engineer in industry (let me have my millennial “back in the day” oilfield story), we picked up a large acreage package in the North DJ Basin that came with an existing DUC. This DUC was critical to underwriting the economics of the acreage, which heightened the criticality of getting the frac design and production profile right. The challenge: no gas infrastructure to support gas lift, rods would not cut it on rate, and we had to embrace ESPs. Trade-offs.

That is a tall order in 5.5-inch monobore within a basin whose best ESP run time to date was something around six months. So I called some really smart people (shout out Jeff Dwiggins) and asked the question you ask when the well economics depend on getting artificial lift right: what is the most reliable system you would put downhole if your career depended on it? The answer, without hesitation, was the RC series. Not because it was the newest or the most technically exotic. Because it was the workhorse. The RC family was the thing you ran when you could not afford to be wrong, the go-to pump that every completions engineer in my network reached for when the wellbore was difficult and the margin for error was thin.

We ran an RC1000. The result: 18 months of continuous run time in a basin where six months was bragging rights. I fully intend to tell my kids about it someday, if they ever ask, which they will not, because I am just the Uber driver.

So when I hear that the RC series, the workhorse of the ESP world, the pump that a generation of engineers defaulted to in their hardest wells, might be getting rationalized, my instinct is to pay attention.

But we do not make commercial decisions on instinct. The whole premise of Kalibr is that anecdotal pattern recognition (”I ran this pump once and it worked great”) is not a strategy. It is a starting point. The question is what signals we can pull from public filings, analyst coverage, and operational data that either corroborate or contradict the rumor. What does the data say that the hallway conversation cannot?

Now I want to be clear. These are rumblings, not press releases. No formal announcement from SLB confirms the explicit discontinuation of specific legacy ESP models.

But oilfield rumors are a lot like the scale in the bathroom. They often tell the truth, even when you would rather they didn’t.

So let’s see what we find.

Learn More About Kalibr

$407 Million Worth of No

Before I show you the financials, I want to say something that may surprise you given the direction of this article: I think SLB’s strategic repositioning is smart. Genuinely smart. Not “smart for the press release” smart. Smart in the way that a well-constructed BATNA is smart, where the move looks obvious only after you understand the positioning it creates.

Here is the thesis. We are in a world of geological headwinds, maintenance CapEx drilling cycles, and PDP-structured entities that need to squeeze every barrel out of every producing well. There are multiple ways to win the next decade in oilfield services. You can build bigger frac fleets and compete on power generation technology, and some companies will win that way. But the capital intensity is brutal and the commoditization cycle is already well advanced. There is another way to win, and it requires a fraction of the capital: embed yourself in the production lifecycle. Every barrel optimized. Every failure predicted before it happens. Every well monitored, chemically treated, and artificially lifted by an integrated system that the operator cannot easily unbundle.

SLB looked at the board and picked that square. From a corporate finance perspective, the math works: production optimization is less capital intensive than drilling and completions hardware, stickier than equipment rental, and, if you can stay ahead of the commoditization cycle, structurally higher margin. The CapEx tells the story. Total capital investments declined to $2.4 billion in 2025 from $2.6 billion in both 2024 and 2023, with management guiding to 5-7% of revenue going forward. You spend less on physical manufacturing. You spend more on software, chemicals, and digital surveillance. The capital intensity drops. The recurring revenue grows. Finance departments love this.

The $4.9 billion ChampionX acquisition (finalized July 2025, all-stock, which is the corporate finance version of paying for dinner by offering to merge your bank accounts) was the enabling move. SLB already held 21% of the global artificial lift market. ChampionX added 5%, plus a production chemicals business, plus the XSPOC optimization platform. Combined: 26% market share and the ability to bundle lift hardware, chemical treatment, and digital monitoring into a single integrated offering that is extremely difficult to unbundle.

Two product catalogs. Two sales organizations. Two overlapping sets of legacy pump lines that were, until recently, competing against each other. You do not maintain both indefinitely. I mean, you *could*. But no one who has ever sat through an integration planning meeting would recommend it, and the bankers who sold you the deal certainly modeled the synergies assuming you would not.

The financial evidence of that rationalization is not subtle. It is, in fact, rather loud if you know where to listen

$407 million in workforce restructuring. $166 million in inventory write-downs. A Production Systems segment where the 12% revenue growth vanishes once you strip out ChampionX and the margin story (”lower profitability in legacy artificial lift,” in management’s carefully hedged language) reads like a quarterly memo from the finance department explaining why the thing they are sunsetting needs to be sunsetted faster. There is a version of the SLB story where the old product lines are a beloved heritage. This is not that version. The incentive to accelerate is not theoretical. It is quarterly.

The replacement products, and the pricing power logic.

This is the part most people will miss, and it is the part that matters most.

SLB is not just swapping old pumps for new pumps. It is doing something more deliberate: introducing products that specifically close the escape routes operators currently use to avoid ESP pricing. This is classic commoditization avoidance, the same playbook you have seen in [frac]() and [compression](), where the vendor introduces differentiated technology that no one else has, commands a pricing premium, and reduces the buyer’s credible alternatives.

Consider what each new product actually competes against.

The Reda Agile compact ESP pushes the pump node deeper into the wellbore and increases drawdown, which makes the ESP option competitive in well profiles where operators were previously concluding that gas lift was the better path. It does not just replace the RC series. It attacks the reason operators were leaving ESPs in the first place.

The Reda PowerEdge ESPCP closes the other exit: the “single run, then convert to rod or gas lift” lifecycle that (full disclosure) I used routinely as an engineer. By extending ESP economics deep into the mature well phase, it eliminates the conversion event that ended the ESP revenue annuity.

And sitting on top of both: the Lift IQ and Lumi platforms, wrapping every pump in AI-driven digital surveillance. SLB’s Digital segment now surpasses $1 billion in annual recurring revenue at 35% accretive margins. The hardware is becoming the delivery mechanism for the software. If you squint, this is less of an oilfield services business and more of a SaaS company that happens to involve things that go downhole.

The sell-side agrees. Unanimously. RBC, Jefferies, Capital One, Bernstein, Piper Sandler, Evercore ISI, TD Cowen, HSBC. Every major coverage name reaches the same conclusion, which is notable given that these are people who are professionally incentivized to disagree with each other. A 26% market share entity with overlapping product portfolios will consolidate. The only question is pace.

Good strategy. Good corporate finance. Genuinely. Some E&Ps will welcome this. If you have been fighting drawdown limitations or burning through ESPs on single-run cycles before converting to rod, the Agile and PowerEdge solve active problems you are already spending money on. Those operators will adopt quickly.

But not every operator thinks linearly inside the ESP category. The ones running tried-and-true legacy systems that work for their wells are not going to see new, unproven technology as an upgrade. They are going to see it as risk. And because they are engineers, not ESP salespeople, they will not ask “which ESP should I run instead?” They will ask a broader question: “Is this the right lift mode at all?” Once you open that aperture, the evaluation is no longer Reda Agile versus RC1000. It is ESP versus gas lift versus rod pump, holistically, with all the upstream implications for casing design, surface equipment, and operating philosophy that come with switching lift modes entirely.

That is where the cascade starts.

Fewer Doors, Same Hallway

If this were just SLB rationalizing a product line, you could shrug. Companies prune catalogs constantly. Operators migrate. Life goes on.

But SLB’s catalog is narrowing inside a supplier market that is simultaneously consolidating at the top, fracturing at the bottom, and walling itself off digitally at every tier. The operator who decides to re-evaluate lift mode will discover that the hallway has fewer doors than it did two years ago.

The top of the market is a fortress. SLB and Baker Hughes now operate with balance sheets clean enough (0.9x and 0.4x net debt-to-EBITDA, respectively) and revenue diversification deep enough (data center power, geothermal, critical minerals) that they can walk away from low-margin ESP tenders without flinching. Post-ChampionX, SLB manages over 20,000 connected assets through Lift IQ and XSPOC. Once your algorithms manage someone’s drawdown in real-time, the switching cost is no longer the pump. It is the data. These are not desperate suppliers. They are choosing their customers.

The middle is gone. ChampionX was the last standalone, publicly traded artificial lift pure-play. SLB absorbed it. The mid-tier negotiation option that operating teams historically played against the Big 3 no longer exists.

The independents face a tariff wall. Most US-based ESP assemblers source pump stages, motors, and cables from Tianjin and Jiangsu. The “Made in USA” label gets applied at final assembly. The core metallurgy is Chinese. Under current tariff structures, these assemblers have no margin buffer. Halliburton itself has acknowledged that artificial lift is its “largest component of tariffs” and that rewiring the supply chain will take quarters.

And in the Permian, where 80-90% of ESPs are deployed under rental agreements and the supplier captures a replacement revenue annuity every 6 to 9 months, the operator’s primary negotiation lever against all of this is the credible threat of converting to gas lift. Jefferies identifies it explicitly: the gas lift BATNA is what forces ESP pricing concessions. It is the only credible exit most operators have left.

Gas lift requires compression.

The buyer’s negotiation leverage in the ESP market and the compression demand signal are the same mechanism viewed from two angles. As the supplier landscape consolidates and the exits narrow, that BATNA gets exercised more frequently. Not because operators want to switch. Because it is the only credible threat they have left.

Subscribe now

So What Do You Do With This

The cascade does not require every operator to convert. It does not require SLB to formally discontinue a single SKU. It requires only that the aggregate probability of gas lift conversion ticks upward at the margin, across hundreds of wells, across multiple basins, across the next 18 to 24 months, as legacy ESP options narrow and the economics of staying versus switching shift.

The mechanism is the 757. Manufacturer rationalizes legacy product. Operators run existing fleet. Fleet ages out. Some migrate within the new catalog. Others discover that their well conditions push them toward an alternative that lives in an entirely different service category.

High-pressure gas lift requires compression.

For compression service providers:

Start having longer-term conversations with operators who lean heavily on SLB ESPs, particularly in basins with limited gas lift infrastructure. An operator whose ESPs are failing more frequently, whose replacement parts are harder to source, whose OEM is steering them toward a product that does not fit their wellbore geometry is going to start asking about HPGL. When they do, they will need horsepower. The obvious demand drivers are the ones everyone tracks. The subtle ones are where the edge lives.

In the Kalibr Compression platform, this is relatively straightforward. We have every compression site mapped. We can flag ESP-specific operations. Overlay that with workover frequency from our signals and you have a targeting map for conversations that no one else is having yet.

For E&P Operators:

If you are running SLB ESPs today, evaluate the new product lines now, before a failed pump and no replacement parts forces the decision for you. If HPGL is even a potential path, begin the compression procurement conversation early. And if you are using the gas lift BATNA as negotiation leverage against your ESP provider, understand that credibility cuts both ways: the more you signal willingness to convert, the more you should actually be prepared to source the compression.

Go Deeper

The Larger Point

The SLB example is interesting on its own. But it is not really about ESPs.

It is about the difference between having data and interpreting it. The world is not short on data. Every public company files quarterly. Every earnings call is transcribed. Every analyst publishes a note. The data is everywhere. What is scarce is interpretation through a lens that actually matters to the person sitting across the negotiating table.

Here is how we think about market intelligence at Kalibr, and it is the framework that made this article possible.

First, aggregate. And if you know the lens you are interpreting through, you can build datasets that do not exist anywhere else. We did not find a public database of every compression site in the Lower 48 mapped to artificial lift method, workover frequency, and vendor attribution. We built one. Because we knew the commercial questions we needed to answer, and the data infrastructure followed the interpretation, not the other way around. This is the part most people think is hard. It is not. It is plumbing with a thesis.

Second, interpret through the lens of negotiation. Not “what does this data mean for the market” in the abstract. What does it mean for the commercial levers available to a specific operating team sitting across from a specific seller? The $407 million in restructuring charges is interesting to an equity analyst. To an operating team negotiating an ESP contract, it is a signal about product availability timelines, spare parts inventory, and the vendor’s willingness to negotiate on legacy equipment they are actively trying to sunset. Different lens. Different output.

Third, apply it to your counterparty. Then apply it to your peers, because your counterparty’s alternatives are dictated by what your peers are doing. If three operators in your basin are already evaluating gas lift conversion, your ESP vendor knows that. Your negotiation position is not independent of theirs.

Fourth, cascade it outward. Upstream. Downstream. Adjacent service categories. The SLB story started as a product rationalization inside an artificial lift division and ended as a compression demand signal. That connection is invisible if you are only watching your own market. It is obvious if you are watching the full topology.

This is the power of interpreting data through the commercial negotiation lens. Not every signal matters equally. Not every rumor deserves a response. But when you can quantify and qualify the signals that do matter, sort them by impact, and map them to specific actions for specific counterparties, you stop being the operating team that reacts to the cascade after it arrives. You become the one that saw it building.

My son cannot do baseball practice, ride his bike, and watch a movie in one night. He has to choose. And his choice, picking baseball, means his sister gets the TV, which means he is catching the tail end of KPOP Demon Hunters for the hundredth time when he finally walks through the door.

Trade-offs cascade. The question is whether you see it coming.

Subscribe now

Nominally Hedged goes out to a small list. If you know someone running compression negotiations this cycle — or who should be — forward it. That's the whole ask.

It worked. D&C costs fell roughly 25 points on a 2015-indexed basis. Genuinely impressive, especially considering the service sector was extracting every possible margin on the way up.

Then everyone looked up from their completions dashboards and noticed that LOE had moved about five points over the same period. Five. The gap between those two numbers is not explained by the relative difficulty of the problem. It is explained by where the industry decided to point its analytical infrastructure.

The oilfield decided that finding and drilling hydrocarbons was the interesting problem, and then proceeded to solve it with genuine intensity. The fact that nobody built the same infrastructure for running the wells afterward is not a failure of imagination. It is a choice, legible in where the industry pointed its people, its capital, and its analytical resources for fifteen years. Drilling and completions teams got more of all three. Production teams got the same headcount and a bigger footprint.

The Kalibr staffing index makes that choice visible in a single data point: the same production team is now managing twice the operational footprint, with tools still pointed almost exclusively at the D&C side of the ledger. The industry did not just underinvest in LOE intelligence. It underinvested while the problem was compounding.

This is the gap. Compression is where it starts.

The Part Where My Last Piece Matters

Last week’s piece established the structural constraint on your compression vendor’s pricing. The short version: their $/HP/month is not a negotiating position. It is the number their entire asset base gets marked against on a consolidated basis for the equity story. When you ask for a rate reduction, you have not asked for a discount. You have asked them to call their auditors.

That analysis revealed where the flexibility actually lives: off-book accommodations that compensate without touching portfolio valuation. Free months, mobilization credits, escalator caps. The accommodations that never appear in the blended $/HP/month figure they report, but absolutely appear in your total cost of ownership if you know to ask for them.

That was half the problem. Here is the other half, and it has three pieces.

The first issue is performance. Compression is a critical-path service. A unit running at 78% mechanical availability is costing you multiples of whatever you saved on rate. The cheapest compressor is the one that runs. Everyone knows this. The problem is that there is no way to benchmark it. Vendors report mechanical availability in their investor presentations, but those are fleet-wide blended numbers that tell you nothing about the specific equipment class or basin where your wells actually sit. You have no third-party performance data. You have the vendor’s word and whatever your field team has observed firsthand.

The second issue is stickiness. Most operators mentally discount their ability to switch vendors because the demob economics look prohibitive. Downtime windows, new contract minimums, transition logistics, lost production. But consider the transaction from the other direction. Once that unit leaves your pad, your incumbent faces a maintenance capex event, uncertainty about whether the unit redeploys in-basin or gets trucked to another region, and carrying cost if it sits in the yard. Their switching cost is not necessarily smaller than yours. Depending on the unit configuration and current supply/demand balance for that HP class in your basin, it may be materially larger. A new vendor can flex on free months or a rebate structure that offsets your transition cost. Your incumbent knows this. And if you can estimate their redeployment risk by compressor configuration — which Kalibr tracks — you know exactly how much leverage you have before you walk into the room.

The third issue is the unit of analysis. Your invoice says dollars per horsepower per month. But horsepower is not what you are buying. You are buying compression duty: the ability to move gas from your wellhead to your sales meter at the pressure differential and flow rate your reservoir demands. Horsepower is how you get there. It is not the destination.

Nobody evaluates a frac spread on raw horsepower. You evaluate whether it delivers the treatment rate and BHTP your completion design calls for with enough redundancy to pump continuously. The HHP figure matters only insofar as it gets you there. Compression works identically. An oversized unit running at 60% load is not a good deal at any rate. An undersized unit as GOR rises is a production constraint dressed up as a line item. Neither shows up as a problem in the $/HP/month contract. Both show up in your net revenue.

The correct unit of analysis is $/MCF moved — what it actually costs to transport a molecule from your wellhead to the meter. This number integrates equipment sizing, utilization, runtime performance, and contract rate into a single figure you can compare across vendors, basins, and well types. It is also the number that nobody in your organization currently has. Getting there requires matching equipment specs to gas volumes and pressures, correlating runtime by vendor and basin, and having contract timing intelligence that tells you when your negotiating window opens.

That is what Kalibr built.

What Kalibr Compression Intelligence Actually Does

The platform is built on three layers of data that do not currently exist anywhere in a single integrated product. Each layer addresses one of the problems above.

Layer 1 is financial — the pricing benchmark. Every public compression provider discloses consolidated blended metrics: fleet HP, blended $/HP/month, reported utilization. What they do not disclose is how those numbers disaggregate by HP band, basin, fuel type, or contract vintage.

This is not an accident. The consolidated number supports the equity story. The disaggregated number would support your negotiation. So you get earnings calls that reference record pricing and 97%+ utilization without any detail on which equipment classes or basins are driving it.

Kalibr reconstructs that disaggregation. The methodology anchors to the one provider that breaks out HP band revenue in their 10-K and infers the band-level rate structure across the industry. The result is a benchmark matrix: $/HP/month by HP band, basin, vendor, and fuel type, with confidence intervals. This is the foundation — knowing what you should be paying before you walk into the room.

Layer 2 is operational — performance and timing. Kalibr’s proprietary asset dataset matches compression equipment in the field — by HP rating, set date, vendor attribution, and the gas volumes and pressures associated with it — to the producers operating them.

This match enables three things public financials cannot. First, regional runtime correlations by provider: which vendors are delivering mechanical availability above or below their marketed guarantees, by basin, by unit size, in aggregate across Kalibr’s observation set. This is the performance benchmark that does not exist anywhere else.

Second, contract timing intelligence: primary term exposure at the vendor level. This tells you when your negotiating window opens and when your counterparty is feeling renewal volume pressure — the information you need to structure flexibility into the deal.

Third, throughput economics. With equipment specs matched to production data, you can calculate actual compression duty and identify where you are paying for capacity that is not moving molecules. This is the $/MCF analysis.

The asset data is what turns financial benchmarking into operational intelligence.

Layer 3 connects vendor financial health to your operational risk. This is the part that functions as a leading indicator — a signal that precedes the outcome by one to three quarters in the historical data.

The logic is straightforward. A vendor deferring maintenance to protect near-term EBITDA margins looks, in the income statement, like they are getting more efficient. Maintenance intensity is falling. Cost per HP is improving. This is the story the equity analyst is writing. It is also the story your account manager is telling you at renewal. What is actually happening is an accumulation of reliability risk that will surface in your operational results before it surfaces in their reported metrics.

The same dynamic applies to labor. A vendor stretching field technicians across more horsepower — HP per technician climbing, disclosed in the filings if you look — looks like productivity improvement. It is actually a response time and preventive maintenance backlog building in the field. The unit on your pad is not getting less attention because it needs less attention.

Kalibr tracks these signals through a composite Reliability Risk Score (RRS) that combines maintenance intensity, labor leverage, and fleet renewal velocity into a single 0–100 metric. A vendor whose RRS is deteriorating is one you want performance guarantees from (or optionality to exit) before the downtime window opens. The financial signal precedes the operational outcome by one to three quarters. That lag is the intelligence.

Layered on top: operational metrics from the asset dataset (HP set intensity, run hours between service, runtime correlations by configuration) that amplify the financial signals with field-level observation. Next week, we will start using this data to benchmark real-time performance in specific basins.

Subscribe now

What This Means for How You Negotiate

Put it together. You have a rate benchmark: $/HP/month by HP band, basin, vendor, and fuel type, so you know whether you are above or below market before you open your mouth. You have contract timing intelligence, which tells you whether your vendor is facing renewal volume pressure in your basin right now. You have a $/MCF analysis, which tells you not whether your rate is competitive but whether your contract is sized correctly for the work it is actually doing. And you have the RRS, which tells you whether the vendor’s forward reliability posture warrants performance guarantees or an exit option.

None of these are rate conversations. Rate is the thing your vendor explained is structurally immovable, and for structural reasons they explained honestly. All of these are contract structure conversations: mobilization credits, performance guarantee provisions, escalator caps, early termination rights. The flexibility exists. It has always existed. The question is whether you showed up knowing where to find it.

The practical application works on three levels.

Benchmark the rate. Know whether your $/HP/month is above or below market for your HP band and basin before you walk into the renewal. This sets the frame.

Evaluate the vendor. Check the RRS. A vendor with deteriorating reliability metrics is a vendor you want performance guarantees from — or a vendor you want optionality to exit.

Right-size the contract. Once you know your actual compression duty, you can identify where you are paying for capacity that is not moving gas. That is not a discount conversation. It is a right-sizing conversation, one your vendor can have without touching their portfolio valuation.

The combination of understanding their model (where the off-book flexibility exists) and understanding your own position (where the optimization opportunities are) is what makes this a better partnership,

Share Kalibr Compression

Kalibr Compression

Starting this quarter, Kalibr is launching the paid tier of Nominally Hedged, anchored by Kalibr Compression, a monthly compression intelligence brief. Each issue covers the following.

Vendor benchmark update. $/HP/month rate movements by HP band and basin across public providers and the independent market. Flagged against prior quarter and trailing four-quarter range.

Reliability Risk Score update. RRS movements for each public vendor, with interpretation of what directional changes mean for operators currently in or approaching renewal.

Contract expiration exposure map. Kalibr’s view of aggregate primary term exposure across the L48 fleet. Regional concentration flags. Where renewal windows are clustering, and which basin supply/demand dynamics are shifting.

Compression intelligence feature. One deep analysis per month. The first issue covers the $/MCF calculation: how to run it against your existing fleet, what the typical range looks like across well types and GOR profiles, and which units to target first.

The free version of Nominally Hedged continues as it has. The paid tier is for operators who want to do something with the analysis.

Leave a comment

The Cost Structure Thesis

D&C intelligence was built for the era when drilling was where the value creation happened. That era delivered 68 on the D&C index against a 2015 baseline of 100. The gains compounded. The margins on the service side compressed. The engineers moved on to the next problem.

LOE is a decade older and largely unchanged, The same systematic intelligence applied to compression, chemicals, artificial lift, and production system design will take it lower; not because production engineers are doing their jobs poorly, but because they are doing those jobs without the tools that D&C built over fifteen years. The operators who close that gap first will have a cost structure advantage that does not depend on commodity prices, inventory quality, or how many rigs their neighbor is running. This is not a cyclical edge. It compounds.

Your vendor has been doing this math since before they picked up the phone. The only question is whether you walk into the next renewal as the operator who brought data, or the operator who confirmed what they already suspected.

Nominally Hedged goes out to a small list. If you know someone running compression negotiations this cycle — or who should be — forward it. That's the whole ask.

Here is a useful theory of the oil and gas industry: it is subsurface real estate development with better PR.

You assemble acreage. You fund development. You sell production at a significant multiple to capital invested, assuming the reservoir cooperates and your completions engineer read a couple of SPE papers at HFTC rather than tapping out the happy hours. This happens underground, which is why most people miss the parallel.

Commercial real estate runs the same play above ground. You assemble land. You fund development. You sell or lease at a multiple to capital invested, assuming the market cooperates and your architect was right about the floor plate. The developer does not own the excavators. They do not own the tower cranes or the concrete batch plants. Just as the developer would never own the equipment moving dirt between the foundation and the first floor, the E&P does not own the compression moving gas between the wellhead and the sales meter. They contract for that capacity, use it, and, in theory, send it back.

The useful part is this: in both cases, the comparative advantage is finding acreage, reading the market, and funding projects. The investors backing that developer, or that E&P, are paying for geological and financial judgment. They are not paying for compressor maintenance margins or crane depreciation economics. So each industry evolves, naturally and correctly, toward a structure where every function is owned by whoever can finance it cheapest.

Now extend the analogy one step further, because this is where it gets interesting.

The compression vendor is not the developer. The compression vendor is the REIT. They own the building. They lease it to you. Their entire business model, the capital structure, the investor base, the return thresholds, the covenant calculations, is built around the assumption that the rent holds. A REIT cannot selectively cut dollars per square foot for one tenant without that conversation immediately becoming a conversation about what every other unit in the portfolio is worth. The auditors ask. The lenders ask. The covenant calculations update. What looked like a one-off accommodation is now a portfolio-wide markdown.

Compression works identically. The dollars per HP per month your vendor quotes is not a negotiating position. It is a valuation input, the number their entire asset base gets marked against. Ask them to cut it, and you have not asked for a discount. You have asked them to reprice their balance sheet.

Which means the first step in any serious compression negotiation is not asking for a lower rate. It is understanding the model well enough to know what they actually can give you, and what the accounting will and will not allow. That requires reconstructing their return thresholds, their leverage math, and the specific utilization floor below which the arithmetic stops working.

What the oilfield supply chain actually is, and almost nobody talks about it this way, is a tranched capital structure wrapped around a set of operational risks. Think of it like a CLO, except instead of slicing credit risk into senior and junior tranches, you are slicing oilfield operational risk into layers and selling each layer to the investors best positioned to hold it. The E&P equity sits at the top of the stack: reservoir risk, commodity price risk, execution risk, in exchange for the IRRs that justify the geological and financial judgment the investors are actually paying for. The compression vendor sits several tranches lower. Contracted cash flows. Hard asset collateral. Multi-year lease terms with investment-grade counterparties. Predictable maintenance cycles. This is, functionally, an infrastructure yield instrument, and it attracts infrastructure yield investors. Pension funds. Insurance companies. The same capital that buys Prologis warehouses and toll road concessions. These investors accept 12 to 15% unlevered returns because the risk profile warrants it and because their liability structures demand duration, not upside.

The tranching is elegant. The tranching is correct. Modigliani and Miller had a theorem about why vertical integration destroys value here: combining high-beta reservoir exposure with low-beta contracted equipment yield produces a cost of capital that is wrong for both businesses simultaneously. The practical version: the E&P that owns its compression fleet has confused its equity investors, annoyed its lenders, and made its CFO’s job harder, all at once.

The ecosystem is not an accident. It is the industry’s solution to a genuine capital allocation problem. And for compression vendors in particular, it has worked almost suspiciously well.

Share

The Part Where the Vendors Start Making Embarrassing Amounts of Money

It has been a good time to own a compressor.

To understand why, you have to start with what happened to the customer. In a previous piece, we traced how E&P consolidation quietly transformed compression from a commodity rental into a critical-path service. The short version: as operators consolidated, pads got bigger, laterals got longer, and simul-frac operations demanded more horsepower delivered to a single location. The architectural decision to run gas lift, driven by capex efficiency and slimmer casing designs, locked compression into the production system the way a conveyor belt gets locked into a factory floor. Rising gas-oil ratios across maturing Permian fields deepened the dependency further. You were no longer renting a compressor. You were building your production system around one.

That shift did two things to vendor economics that do not show up immediately in the headline numbers. First, it upgraded the counterparty. The operators committing to compression contracts on large centralized facilities are not small privates running a three-well pad on a prayer. They are Devon and Diamondback and Coterra, locking in compression for drilling programs that have already been AFE’d, approved, and scheduled. Investment-grade credit risk on mission-critical equipment. The contract book got longer, the credits got better, and the embedded switching costs got higher, all at the same time.

Second, and this is the part that rarely gets written about: the supply chain collapse did not just constrain new capacity. It restructured how vendors carry risk. Caterpillar engine lead times stretched to 90 to 120 weeks at the peak, call it two years from order to delivery on a large horsepower unit. Vendors responded by requiring operators to commit to contracts before equipment was ordered. The mechanics matter here. The vendor does not pay the balloon payment to Caterpillar or the engine manufacturer until the unit sets on location, which is also when the contract starts and revenue begins. No float. No inventory carrying cost. No speculative deployment. The vendor is running a pre-sold order book against a manufacturer who holds the inventory risk.

The numbers are what you get when all three of those things compound simultaneously. USAC reported revenue per horsepower per month of $21.69 in Q4 2025, a record. Archrock’s operating margins have exceeded 70% for five consecutive quarters. Kodiak’s fleet utilization sits at 97.7%, which in a capital-intensive equipment business is essentially the speed of light. You do not get meaningfully closer to full without something breaking. USAC has made 50 consecutive quarterly distributions without a cut since its IPO, returning approximately $1.9 billion to unitholders.

The bad debt figure is the one that stops the room. Over nineteen years, USAC’s write-offs have totaled 0.06% of billings. Not 6%. Zero point zero six. When your customers are investment-grade operators running production systems they cannot shut down, and your contracts were pre-sold before the equipment existed, bad debt becomes a rounding error on a rounding error.

To put that in context your CFO will appreciate: USAC’s three-year total shareholder return of 54% nearly matched the S&P 500 Value index during one of the most aggressive tech bull markets in history. Compression does not have a ChatGPT moment or a Blackwell chip launch. It has a Caterpillar 3608, a multi-year contract, and a customer who cannot turn off the gas lift without killing the well. It turns out that is a durable competitive position.

Which makes the Axip Energy Services bankruptcy filing, Southern District of Texas, February 22, 2026, genuinely interesting in a way that has nothing to do with sympathy for Energy Spectrum Capital’s limited partners.

In a market this favorable, going broke requires some effort. But here is the thing about a compression bankruptcy in a compression bull market: it is the only controlled experiment you are going to get. Every healthy vendor is running the same fundamental model, same return thresholds, same accounting logic, same structural constraints. But they are not required to show their work. Axip is. The capital structure, the return thresholds, the utilization floor below which the arithmetic stops working, the exact sequence of events that made refinancing impossible, all of it goes into the public record, sworn and filed under penalty of perjury. Most operators will read the headline and move on. That is the wrong read. The Axip filing is a tutorial on the model your vendor is running, and therefore a precise map of where the leverage actually sits when you are sitting across the table from Archrock or USAC or Kodiak at contract renewal.

Share Kalibr - Strategy Research for Oil & Gas

The Part Where We Pull Out a Calculator

On February 22, 2026, Axip Energy Services filed for Chapter 11 in the Southern District of Texas. Energy Spectrum Capital had acquired the company in September 2022, loaded it with $240.5 million in debt, and watched the thesis fall apart in roughly three years. The stalking horse bid from Service Compression LLC came in at $161 million, 33 cents on the debt dollar.

The capital markets press wrote it up as a private equity cautionary tale. That reading is not wrong. It is just not useful.

What the filing actually is, the First Day Declaration, the capital structure schedules, the forbearance amendments, all of it sworn under penalty of perjury, is a detailed post-mortem on what the occupancy model looks like when it breaks. The REIT parallel from earlier is not decorative here. Axip did not fail because they were bad operators. They failed because vacancy crossed a threshold their debt structure could not survive. Understanding exactly how that happened tells you more about your vendor’s negotiating constraints than anything they will voluntarily disclose across a contract renewal table.

The building and the mortgage

Axip operated 940 units totaling 326,070 HP across seven facilities in Texas, New Mexico, North Dakota, and the Gulf of Mexico. Against that asset base sat $240.5 million in funded debt: an ABL facility with JPMorgan Chase at $207.8 million, a superpriority tranche at $13.2 million, and a second-lien held by Permico Inc. at $19.5 million. At roughly $92 million in annual revenue, that is a 2.6x debt-to-revenue ratio. There is no margin for error in that structure. A building owner running 2.6x debt-to-rent does not survive a vacancy spike. Neither does a compression vendor.

Note the second-lien holder. Permico is PE-affiliated. Seller financing at the junior layer is a signal that the leverage could not be fully placed with third-party lenders at closing. It is not conclusive, but it is a tell worth filing away when you are evaluating vendor financial health.

The tenant that did not pay and could not leave

In Q1 2024, a major Gulf of Mexico customer filed Chapter 11 and converted to Chapter 7 liquidation. Twenty-four offshore units, over 15% of Axip’s total horsepower, were stranded overnight. The customer did not pay demobilization costs. The units could not be economically redeployed onshore. Offshore compression does not truck to the Permian. It sits.

The EBITDA impact was immediate and unrecoverable. But the more important lesson is structural: one counterparty event erased 15% of revenue on a balance sheet with no cushion to absorb it. This is the right-tail risk that the occupancy model carries silently until it does not. For operators evaluating vendors, counterparty concentration is not an abstract risk factor buried in a 10-K. It is the variable that determines how much negotiating room your vendor actually has, because a vendor carrying fragile customer exposure needs your stable, investment-grade contract more than their renewal posture suggests.

The building that needed a generator to turn the lights on

More than 25% of Axip’s fleet was electric-motor driven. On paper, a forward-looking strategic position. In practice, a mismatch between strategy and ecosystem.

Electrification is not a bad bet. Archrock and Kodiak are making similar investments, and the long-term direction of the market is not seriously in dispute. But executing an electrification strategy requires your customer base to be able to support it: reliable grid access, infrastructure buildout, power availability that matches compression demand. Axip’s customer mix could not deliver that. When grid availability failed to keep pace with deployment, customers ran generator sets to power units they were already renting, effectively paying twice for the same compression. That math accelerates the decision to return equipment considerably.

The contrast with Archrock is instructive. Archrock can pursue electrification credibly because their customer base includes operators large enough to command grid infrastructure buildout. ExxonMobil does not run generator sets. Axip’s customers did. Strategy requires ecosystem alignment, and a vendor pursuing a strategy their customer base cannot support is a vendor burning capital and absorbing returns. Which is both a risk signal and, if you are in their customer base, a leverage signal.

What the vacancy actually cost

Axip’s pre-stranding revenue ran approximately $92 million annually against a fleet of 326,070 HP, an implied rate of $23.52 per HP per month across the total fleet. But 15% of that fleet was stranded and generating zero revenue. Calculate on deployed horsepower only and the effective rate was $27.67 per HP per month, higher than USAC’s record $21.69. Axip was not losing the rate war. The business model was working on the equipment that was working. What it could not survive was the vacancy.

Post-stranding, revenue dropped to approximately $78 million. Gross profit at industry-standard margins fell to roughly $51 million. Debt service alone consumed most of that. The remaining operating cost structure consumed the rest. There was no path to solvency that did not require either a significant equity injection or a buyer willing to reset the capital structure, and after six months and 85 parties contacted in an Evercore-run process, neither materialized.

What your vendor’s balance sheet is actually telling you

The Axip autopsy is not a story about bad luck or bad management. It is a demonstration of how the occupancy model fails, and therefore a precise illustration of what makes it fragile even when it appears to be working.

Three things to take away.

First, utilization is not a performance metric. It is a solvency metric. For a vendor running 4x leverage against contracted equipment cash flows, the difference between 95% and 80% utilization is the difference between distributable cash and covenant breach. The problem is that in the current market, all three majors are running 94 to 98%. The headline number does not tell you much. What matters is primary term exposure: which contracts are rolling, when, and what that does to forward utilization when the aggregate looks healthy. That is where the fragility hides. It is also where the negotiating window opens, if you know where to look. (Kalibr tracks primary term exposure at the vendor level. More on how to use that in a future piece.)

Second, counterparty concentration is a commodity exposure signal, not just a default risk flag. Consolidation has largely removed outright bankruptcy risk from the E&P customer base. Devon and Diamondback are not going Chapter 7. But concentration still matters, because it tells you what your vendor’s growth outlook is actually tied to. A vendor with a gas-heavy contract book in a weak gas price environment is a vendor whose forward EBITDA is under pressure. If you are an oil-weighted operator in the Midland Basin, you are not just a renewal. You are a hedge against the exposure that is compressing their growth outlook. That asymmetry is leverage, and it is leverage most operators never think to calculate because they are focused on their own commodity exposure, not their vendor’s. Read their 10-K customer concentration disclosures and map it against the current price environment before you walk into a renewal.

Third, know your vendor’s strategic white space and know where you sit inside it. Every vendor has edges where they are strong and edges where they are exposed. A vendor overextended into a technology or geography their customer base cannot support is a vendor absorbing execution risk that will eventually show up in your service quality or their willingness to deal. Understanding the shape of their strategy tells you where the pressure points are before they become your problem.

The practical starting point for all three is the same: model the occupancy floor. What utilization does your vendor need to service their debt at current rates, and how close are they to it? That calculation sets the realistic bands of negotiation, not what you want, but what the arithmetic will actually allow. The inputs are their cost basis, their WACC, their stated IRR targets, and their leverage structure. Archrock, USAC, and Kodiak publish most of this. What they do not publish is reconstructible from public filings. (This is what Kalibr’s market intelligence package is built around: the vendor-level economics that turn a renewal conversation from a negotiation about numbers into a negotiation about constraints.)

Once you have the floor, the last piece is understanding what form a concession can actually take. Rate cuts are a balance sheet conversation your vendor cannot have without repricing their entire portfolio. The auditors ask, the lenders ask, the covenant calculations update. But off-book accommodations are a different category entirely. Free months, mobilization credits, escalator caps, priority dispatch commitments: accounting-friendly concessions that do not touch the portfolio valuation. Understanding which category each ask falls into is the difference between a negotiation that goes somewhere and one that does not. That is where we are going next.

Subscribe now

The Part Where We Tell You What to Do With All of This

Here is what you now know that most operators sitting across a compression renewal table do not.

The rate your vendor quoted is not a negotiating position. It is the number their entire asset base gets marked against. Ask them to cut it and you have not asked for a discount. You have asked them to call their auditors and explain why their portfolio is worth less than it was yesterday. That conversation does not happen. Not because your vendor is unreasonable. Because their accounting will not allow it. The REIT does not discount one floor of the building. The compression vendor does not discount one contract. The math is the same.

What the Axip filing gave you, and what three years of record utilization, record margins, and record distributions from the healthy vendors obscures, is a precise picture of how fragile the occupancy model actually is. One counterparty event. Fifteen percent vacancy. A debt structure with no cushion. The building empties faster than the mortgage adjusts. That fragility does not disappear because USAC is at 94.5% utilization and Archrock’s dividend coverage is 4.9 times. It goes underground. It waits. And it surfaces, quietly, in the primary term exposure of a contract book that nobody outside the vendor’s treasury department is tracking.

Until now, anyway.

Build the model. It does not have to be perfect. It has to be better than walking in blind, which is a low bar to clear. Use their public filings. Use their stated IRR targets and leverage ratios and cost of capital disclosures. Back into the utilization floor below which the arithmetic stops working. Map their counterparty concentration against the current commodity price environment and ask yourself whose problem you are solving by signing a renewal. That is your band. That is the realistic range of what this negotiation can produce, before anyone has said a word about rates.

Then, and this is the part that turns a framework into a result, look for the intelligence that tells you where you sit inside that band. Primary term exposure. Counterparty concentration. Strategic overextension. The signals that tell you whether you are negotiating from the middle of their comfort zone or the edge of it. (Kalibr tracks this at the vendor level, if you would rather not spend a weekend with a Bloomberg terminal and three years of 10-K filings. Either way, the inputs exist.)

What you will find, once you have done this work, is that the negotiation looks different than it did before. Not because the vendor suddenly became generous. Because you stopped asking for things the accounting will not allow and started asking for things it will. There is a whole category of concessions that compress vendor economics without touching the portfolio valuation, concessions that are, in some cases, worth more to an E&P operation than a rate cut would have been anyway.

That is next week.

But here is the thing to sit with until then. Every sophisticated commercial real estate tenant knows that you do not walk into a lease renewal and ask the landlord to reprice the building. You ask for three months free rent, a tenant improvement allowance, and a right of first refusal on the suite next door. The landlord says yes, the portfolio valuation stays intact, and both parties leave the table having gotten something real. The dollars per square foot never moved. Nobody’s auditors had an uncomfortable morning.

Your compression vendor is the landlord. You have been asking them to reprice the building. You need to be asking about the tenant improvement allowance. More to come.

Nominally Hedged goes out to a small list. If you know someone running compression negotiations this cycle — or who should be — forward it. That's the whole ask.

Let’s say a foreign company makes widgets (Yes, I love widget examples). Congress says that’s a national emergency. Three agencies get to work.

Tool one: the antidumping duty. The Commerce Department figures out what the widget should cost in the foreign producer’s home market, then turns that theoretical cost advantage into a tariff rate. Some came in above 100%. Others at 44.93%. Others at 78.30%. These numbers imply very careful math. Whether they correspond to anything observable in the real world is a question the lawyers prefer not to examine, because the lawyers get paid either way. The widget does not have an opinion.

Tool two: the countervailing duty. Same idea, different offense; not “you’re selling too cheap” but “your government is helping you sell too cheap.” Stacked directly on top of the first rate. Two agencies. Two rate-setting processes. Two sets of lawyers in no hurry.

Tool three: Section 232. Different game entirely. The first two tools are about fairness. This one is about national security. A blanket 25% tariff, no dumping finding required, no subsidy calculation necessary. Security matter, done. The widget is now a geopolitical actor.

All three stack. The widget isn’t just expensive. It’s expensive in three directions at once, for three official reasons, administered by three agencies who have never been in the same room and have no intention of changing that.

Then the exceptions. Bilateral carve-outs, hard quotas, tariff-rate agreements; each country negotiating its own arrangement, each deal its own legal instrument, each instrument its own lawyers. One producer negotiated an import allowance calibrated to an exact ton (not a round number, an exact figure) in a document justified by national security. The widget, at this point, probably has its own trade attorney and a LinkedIn post highlighting “The 10 things I learned about B2B sales through tariffs.”

By the mid-2020s, the machine had issued so many bespoke exceptions that it had stopped sorting anything. Then the reset. Every exemption revoked. Blanket tariffs reinstated and doubled. Years of learned behavior, erased overnight.

The widget doesn’t care what you call it. It does care that it is OCTG, and it goes in the ground.

Subscribe now

One Market, Then Two

A stacked tariff doesn’t hurt all pipe equally. It hits premium and commodity grades asymmetrically, and that asymmetry, without a single phone call or industry meeting, sorted an entire market.

At the premium end, the purchase decision was never just about price. There’s a specification, a qualification process, a technical relationship, institutional knowledge built over years. A foreign producer competing here isn’t fighting the tariff alone. They’re fighting the tariff plus the qualification barrier plus the certification cycle plus the accumulated weight of every prior purchase decision. Domestic producers looked at this and made a rational choice: this is where we live. The tariff didn’t just protect premium, it calcified it. Producers with strong balance sheets, domestic manufacturing, and proprietary connections were going to win this segment regardless. The tariff just meant winning it required less effort, at higher margins, against a field that had been administratively thinned.

The commodity end is a spreadsheet. No proprietary spec. No qualification cycle. Just: what does it cost delivered. The tariff bites here too (50% is still 50%) but a foreign producer with structurally lower input costs can absorb enough to stay competitive. Domestic producers ran the same calculation in reverse: why fight a cost-per-unit battle against someone with lower labor and energy costs when the premium end is sitting there uncontested?

Nobody coordinated this. Everyone read the same signal, the tariff gradient, permanent and public, and independently arrived at the same answer. Domestic producers migrated up. Foreign producers concentrated down. One market became two. The government spent fifteen years building a machine to protect domestic producers and accidentally built the market structure that a McKinsey team would charge seven figures to recommend.

McKinsey would call it strategic portfolio optimization. The game theorists would call it a coordination equilibrium sustained by a common focal point. The lawyers would call it Tuesday.

The operations group that needs to catch up is on the buy side, because the old mental model, domestic and import mills competing across the full product spectrum, price tension everywhere, a credible alternative always one call away, is simply no longer accurate. The segmentation is real, structural, and built on a tariff schedule that is not going away. Most operations playbooks were written for the market that existed before it was reorganized.

You Brought a Weathervane

When operations groups use data to make OCTG decisions, it tends to come from two places. Neither was designed for the market that currently exists.

The rig count model. The industry default demand framework, and it was reasonable when it was built. Rig count as a demand proxy fails to capture drilling efficiency gains — fewer rigs are drilling more footage than three years ago, which means rig count systematically understates demand intensity as that trend continues. It cannot account for lateral length inflation, which changes not just the volume of OCTG demanded but the specific product mix. And it has no mechanism for the bifurcated operator environment you’re actually in right now, where a well-capitalized major on a board-approved program and a leverage-constrained PE-backed private respond to the same WTI print with completely different drilling behavior.

A model that cannot distinguish those two cases is not a demand forecast. It is an activity count. It tells you how many rigs are running. It does not tell you what they are buying, from whom, or at what point in the product stack the pressure is actually building.

Pipe Logix. This one deserves careful treatment, because it is widely cited and widely misunderstood — a combination that tends to be expensive for whoever is on the wrong side of it.

Pipe Logix is a distributor sentiment survey. A diffusion index constructed from voluntary responses by a panel of distributors on questions about order books, inventory, and price expectations. The benchmark pricing data is a legitimate spot market reference. The problem is not the index itself. The problem is what happens when you build a term contract around it.

The whole logic of committing volume to a mill is straightforward: you trade flexibility for price certainty, the mill trades margin for utilization predictability, both sides capture something real. That trade only works if the price you lock in actually reflects the market at the time you commit. Pipe Logix indexing quietly voids that logic. Changes take one to two quarters to flow through into contract pricing — Tenaris management confirmed this publicly regarding their own North American contract reset cadence. The index declined four consecutive months through December 2025. An operations group on an indexing agreement in that environment continues paying near-peak rates for six months after the physical market has already corrected.

What that means structurally: you have surrendered the volume commitment, which is real and binding, in exchange for price certainty, which the indexing lag has made illusory. You have locked in the obligation without locking in the benefit. At that point a spot buyer, no commitment, full flexibility, is actually better positioned to capture the market as it moves. The index doesn’t make term contracts bad. It makes indexed term contracts a mechanism for giving up optionality without getting paid for it.

There are two additional construction issues worth noting, though neither is the primary concern. Survey respondents report prices with an inherent incentive to cluster near prior month levels — sharp declines devalue their own inventory — so with distributor sentiment at lows and roughly three months of supply in the market, the index likely overstates true clearing prices. And the seamless segment has actively decoupled from Pipe Logix when the market falls while using it as a floor when the market rises, a dynamic JP Morgan has specifically noted in Tenaris pricing commentary. Both are real. But the more fundamental issue is the one above: the indexing lag structurally inverts the trade you thought you were making.

Pipe Logix is a weathervane. It tells you which way the wind is blowing after it has already arrived. A weathervane is the right tool until the weather system changes. When it does, you need a forecast.

Subscribe now

The Number Under The Number

The right starting point is not what an index says. It is what it actually costs to make steel tube and put a connection on the end of it. This is harder than it sounds, and the difficulty is not accidental, cost opacity is a competitive advantage for the people selling you pipe.

The domestic OCTG market runs on two fundamentally different steelmaking processes with completely different cost structures.

The EAF mill melts recycled scrap using electricity. Primary input costs are two variables: scrap and power. Both are volatile. Both trade on liquid futures that encode the market’s real-time view of where input costs are heading — which means a disciplined analyst can build an EAF cost floor from public data alone, updated monthly, without a single conversation with the mill. The yield dynamics of the seamless piercing process introduce a cost multiplier that most surface-level analyses get wrong, and getting it wrong compounds into meaningful floor estimation error.

The ERW mill starts with hot-rolled coil, forms it into a cylinder, and welds the seam. Better yield. Lower energy intensity. Lower capital requirement. Structurally lower cost floor. ERW and seamless are not the same product competing on the same cost basis. A procurement strategy that treats them as interchangeable is leaving money on the table before the conversation starts.

Then there is the third archetype that most cost analyses miss entirely: the billet importer.

Some domestic mills do not produce their own substrate. They import semi-finished steel, billets or green tubes, and run them through downstream operations domestically. The finished product is technically domestic pipe. The substrate carries the full Section 232 tariff on the way in.

At a 50% Section 232 rate, imported billet adds a meaningful per-ton penalty relative to a domestic EAF competitor sourcing scrap at market. The producer who built their U.S. manufacturing footprint around imported substrate is paying the tariff that was ostensibly designed to protect them. The wall keeps out the competition and simultaneously taxes the raw material. This is the kind of outcome that makes trade lawyers wealthy and procurement analysts confused.

The practical consequence for a buyer: the billet-import producer has extraordinary competitive strength in the premium segment — proprietary connections, rig-direct distribution, deep operator relationships — and a structurally challenged cost position on commodity grades. They are not trying to win commodity casing. Their cost structure is one of several reasons why. Which changes the bundling conversation considerably.

A few layers that matter on top of raw substrate:

• P110 heat treatment. Quench and tempering adds meaningfully to the cost structure.

• Threading and connections. This is where the cost stack becomes a hierarchy. Fully integrated seamless with proprietary in-house connection technology and rig-direct distribution captures the highest revenue per ton and carries the deepest moat. The connection is not just a product — it is a switching cost embedded in every well program you run. Understanding that is the first step toward negotiating it correctly.

• Freight. Systematically underweighted. Mills were sited deliberately for their basins. A Haynesville ERW mill sitting inside the basin has already won on freight to Haynesville before your RFQ goes out. Procurement strategy that ignores geography is fighting a structural advantage instead of using it.

The one variable that is not directly observable — and the one that matters most when the market is turning — is utilization. Fixed cost absorption moves materially with utilization, and it is the variable mills have the most incentive to obscure in public commentary. The difference between a mill at 70% and 90% utilization represents meaningfully different cost floors, and that difference determines whether current market prices represent a floor or a ceiling with room to fall.

There are observable signals that triangulate utilization without requiring the mill to disclose it. Mapping those signals changes the procurement conversation from what does the index say to how close is the market price to the point where this specific producer starts curtailing. Those are different questions. The second one is the right one.

What The Mill Thinks Of You

The supply-side cost floor tells you the bottom. It does not tell you when the market gets there, how long it stays, or what your specific buying power looks like relative to everyone else competing for the same pipe.

The conventional demand model asks: how much pipe does the market need? That is a reasonable question and also an incomplete one. The market is not a single buyer. It is approximately 650 operators across fourteen basins with different capital structures, different board mandates, different hedging books, and fundamentally different propensities to drill when WTI prints the same number.

A model that starts with aggregate rig count cannot distinguish a well-capitalized major on a board-approved five-year program from a leverage-constrained PE-backed private on a borrowing base renewal. Those two operators respond to the same commodity price signal with completely different behavior. Conflating them is how aggregate forecasts systematically overstate demand at the tails and miss inflections at the turns.

The insight that most operations groups have never formalized: your position on the demand curve is not just a function of how much pipe you buy. It is a function of how reliably you buy it, how predictably you can commit to future volumes, and how that reliability compares to every other buyer the mill is evaluating simultaneously.

A large operator with variable drilling behavior is less valuable to a mill than a smaller operator with a credible multi-year commitment — because the mill is not just selling pipe, it is managing utilization, and utilization predictability has real economic value that shows up in pricing. The operator who understands this can negotiate it. The one who doesn’t pays the price of being treated like a spot buyer even when they’re not.

Where you sit on the volume-conviction surface determines not just what discount you should be able to negotiate, but which mills have the most incentive to compete for your business, at what point in their fiscal calendar, and under what contract structure. Most operators have never mapped where they actually sit. The ones who have are having different conversations with their vendors.

The Midland and Delaware Basins together account for the dominant share of 5.5” P110 production casing demand. The top two specs represent a disproportionate share of total footage demand — both concentrated by basin, with bear-to-bull sensitivity in the 10–13% range. That figure is the model’s answer to a question most operations groups haven’t asked: how much does my buying environment change between a $57 WTI world and a $72 WTI world? The answer, at the spec level, is more than most contracting postures have priced in.

Distributor inventory is elevated. Working capital pressure is visible in publicly disclosed financial metrics for anyone paying attention. The difference between an operations group that understands the distributor’s financial position and one that doesn’t runs $100–200 per ton on identical pipe. That gap is not permanent — it closes as the supply overhang works off. The cost model tells you where the floor is. The demand model tells you when the window closes.

Together they answer the one question operations groups actually need answered: lock in now, or wait?

Timing Isn’t Neutral

The rig count model and the Pipe Logix index were fit for purpose when the market they were built to describe actually existed. The market reorganized. Nobody updated the instruments.

The operations group that has mapped its own position on the demand curve, built its own cost floor, and walked into a renewal knowing what the mill’s utilization-adjusted breakeven looks like — that group is having a negotiation. Everyone else is having a pricing conversation where only one side brought data.

The lock-in window is open. How long it stays open is a function of inputs the rig count cannot see. When it closes, it will close faster than the index confirms it.

Ignorance is not neutral. It is a pricing signal the market reads every time you sign a renewal.

This is the first in a series. The framework above is the foundation; supply-side cost architecture, demand conviction, and the analytical connective tissue that actually answers a contracting question. The interesting part starts now.

A few of what we’ll work through in the coming weeks:

What actually happens to demand when two major operators consolidate? The press release says “synergies.” The demand model says something more specific: which specs compress, in which basins, on which timeline, and what that means for the mills who had both operators as customers and now have one.

If Pipe Logix is a structurally compromised indexing mechanism (and we’ve made the case that it is) what does a better one look like? There are observable market inputs that encode real clearing price information without the survey bias, the inventory incentive distortion, or the one-to-two quarter lag. We’ll build one.

What happens to 4.5” P110 demand concentration when a major Bakken operator pauses its drilling program? The spec is already highly concentrated by basin. A single operator decision can move the supply-demand balance materially — and the pricing response may arrive before most buyers realize the shift has happened.

These aren’t hypotheticals. They’re the questions that fall out naturally when the model is granular enough to ask them. The rig count can’t answer any of them.

Same time next week. Bring your AFE.

Then it gets harder. Your tool is not that difficult to replicate, and your client base consolidates faster than your competitors exit, which means more buying power on one side of the table and a price war developing on the other. You respond the way every serious operator responds: tighter geography, vertical integration, proprietary systems, performance contracts. These work. Mostly. But you are still selling a commodity, and commodity sellers eventually compete on price, and competing on price is a terrible way to spend a career.

So you build a better tool. Not incrementally better — fundamentally different. A tool that nobody else has yet, that demonstrably cuts your client’s cost per job, and that requires serious capital commitment on your part to deploy. Here is where it gets interesting: your client is playing a completely different game. They are a cash flow maximization program. They have a board expecting returns, a buyback target to hit, and a CFO who measures success in dollars returned to shareholders, not equipment purchased. They love the story of cost reduction. They will put it in their investor deck. What they will not do is write the check to make it happen. So when you show up with the better tool and ask for term in exchange for making the capital commitment yourself, they are not reluctantly agreeing. They are relieved. The term contract is not a concession — it is the product. You get EBITDA predictability and the investment thesis to deploy capital. They get the technology, the narrative, and none of the depreciation. Everyone shakes hands and actually means it for once.

The problem is the old tool. You still have a fleet of them. In a soft market, clients ask for the old tool and make you compete on price, which pulls your blended margin back toward a number your CFO mentions at every ops review with the same face. You cannot unilaterally retire them without ceding share to competitors who still offer them. You cannot call your competitors and agree to retire the old tools together, because this is fundamentally not how capitalism is supposed to work. Lina Khan may no longer be running the FTC, but the Sherman Act did not leave with her.

What you can do is make a series of carefully worded public statements — in earnings calls, in investor decks, in the particular register of language that sounds like operational housekeeping but lands as strategy — indicating that you are not merely taking the old tools out of service. You are dismantling them. Parting them out. Making the retirement structurally irreversible, not just temporarily inconvenient. Your competitors are listening, because they have the same problem and the same CFO giving them the same look. They may reach the same conclusion independently. The old tools disappear from the market without anyone having to say anything that a regulator would find interesting.

This is, roughly, what happened to conventional diesel frac fleets over the course of 2025. The Q4 earnings cycle just told us whether it worked. The short answer is yes. The more interesting answer is that “it worked” created a follow-on problem that now belongs entirely to E&P operators: where the old market had one commodity tier and some premium options, the new market has three distinct next-generation architectures with different economics, different operator fit, and very different responses to a buyer who shows up without a credible alternative. Someone has to understand what they are. It might as well be you.

Subscribe now

1.1 Million Horsepower Walked Out the Door and Nobody Is Replacing It

Let’s do the ledger first, because the numbers are specific enough to matter.

The D&C index sits at 75 — down from 100 in 2015 — while the staffing index has doubled to 18. Operators are completing more wells per crew, returning more capital per barrel, and running programs explicitly designed to outperform at $61/bbl. That backdrop matters because it explains the behavior on the service side. When your client base is structurally committed to doing more with less, the rational response is to concentrate capital where the returns are highest and liquidate everything else. That is exactly what happened.

Over 2024 and 2025, the top five pressure pumping providers retired, impaired, or scrapped more than 1.1 million hydraulic horsepower of conventional diesel equipment. Patterson-UTI removed 600,000 HHP over two years. ACDC retired 400,000 HHP in a single quarter. ProPetro took a $189 million write-down and described the remaining diesel fleet as being “harvested for cash” with zero maintenance CAPEX. Halliburton described their conventional idle capacity as “consciously stacked and price-dated,” which is a slightly more collegial way of saying the same thing. Liberty Energy cannibalized their legacy equipment for spare parts. In no case was any of this equipment stacked while awaiting a better market (with the exception of HAL). It was stripped for transmissions and power ends, written off, and in several cases physically scrapped.

Zero dollars replaced it in kind. Not one provider discussed building new Tier 4 DGB equipment on any earnings call across eight consecutive quarters. The existing Tier 4 fleet is the fleet. Minus annual attrition, which nobody is replacing either.

This does not mean Tier 2 and Tier 4 DGB equipment vanished. It means they became something different: the new commodity floor. The old commodity tier was Tier 2 diesel. The new commodity tier is Tier 4 DGB — the equipment that every provider still has in some quantity, that sets the baseline market price, and that service companies are increasingly trying to deploy in simul-frac configurations rather than single-well operations. The reason is straightforward: running two wells simultaneously with Tier 4 DGB requires more horsepower per location, reduces crew count relative to stages completed, and improves service company margin per job without requiring newbuild capital. Your vendor’s enthusiasm for getting you onto simul-frac with their conventional fleet is not purely about your cost structure. It is also very much about theirs.

Every dollar of newbuild capital went above the commodity tier. And the thing about next-generation equipment in frac is that it is not one thing — it is three distinct things, each with a different power source, a different capital structure, and a different answer to the question of which operators it actually fits.

The Part Where We Explain That “Electric Fleet” Is Not One Fleet

The conventional framing in E&P operations is: there are old diesel fleets, and there are new electric fleets, and you should try to get the new electric fleets because they are better. This framing is approximately true the way “a plane and a helicopter are both aircraft” is approximately true. Useful at 30,000 feet. Unhelpful when you are trying to figure out which one lands in your backyard.

Every dollar of newbuild capital is flowing into one of three distinct architectures. They share a fuel source and a general direction of travel. They do not share a capital cost, an operating envelope, or an operator fit.

HAL Zeus runs aircraft-derivative gas turbines generating ~35MW of continuous electrical output distributed across purpose-built electric pumps. At approximately $60 million per fleet it is the most capital-intensive deployment in the peer group — and build-to-contract only, meaning no spot market exists. The operational payoff is real: Zeus executes simul-frac at ~65,000 HHP versus 100,000 to 120,000 HHP for legacy configurations running the same job design. The switching cost is structural. Zeus deployments embed wellsite power infrastructure, data integration, and choreography that is incompatible with any competitor fleet. Operators who built programs around Zeus efficiency metrics have created a durable relationship with Halliburton whether or not they intended to.

LBRT digiFrac and digiPrime are two distinct platforms that Liberty’s investor materials tend to flatten into a single “natural gas fleet” narrative. digiFrac is a clean-sheet purpose-built electric pump with the best utilization record in the peer group: 7,143 pumping hours per year on a single fleet, representing 96% of the calendar (~19.5 effective hours per day) at operating pressures exceeding 10,000 psi. digiPrime is mechanically different — a variable-speed natural gas reciprocating engine in direct mechanical coupling, eliminating the electrical conversion step. Lower capital cost, simpler infrastructure, and a program fit that does not require Permian-scale pad operations to justify the economics.

PTEN Emerald takes the mechanical drive premise further still. No turbine, no electrical system, no conversion losses — natural gas combustion to pump motion in a single step. Capital runs 25 to 30 percent below a full e-fleet at 100% diesel displacement. The differentiation from Zeus and digiFrac is program fit: Emerald does not require the water logistics of trimul-frac (~330,000 bbl/day versus ~110,000 for a conventional zipper) and works in basins and completion designs where pad scale cannot support electric fleet economics. You change the iron. You do not change the program.

The Part Where We Pull Out a Calculator

The conventional way to track frac market tightness is total horsepower. Count the active fleets, multiply by nameplate HHP, compare to demand. This number is now systematically misleading, and the operators still using it are making procurement decisions with a broken instrument.

Total horsepower deployed across the peer group remained roughly consistent through 2025 even as fleet counts declined. The reason: simul-frac and high-efficiency operations require 30 to 100 percent more horsepower per job than conventional single-well pumping. A market that looks flat in raw HHP terms is actually tightening in job capacity. The retired Tier 2 diesel was completing approximately 60 percent of the stages per month of a next-generation fleet. When you retire 1.1 million HHP running at 60 percent efficiency and replace part of it with equipment running at 95 percent, the aggregate horsepower number goes down but the job capacity does not. What also does not survive the transition is the pricing that attached to the old equipment.

The E&P operations engineer benchmarking their 2026 frac budget against 2023 actuals is comparing against a market that no longer physically exists. The equipment that set those prices has been parted out. What replaced it has a different cost structure, different margin expectations, and a very different answer when a buyer shows up without a credible alternative.

Q4 2025 Earnings: What the Numbers Actually Said

The Q4 2025 earnings cycle is the most useful dataset this peer group has produced in several years, for a straightforward reason: soft markets make financial disclosures honest. The narrative in prepared remarks — “steady,” “resilient,” “disciplined” — runs about two minutes before the analyst Q&A begins disassembling it. The job is to read the transcripts, the 10-Ks, and the broker notes together, and build a vendor picture based on what the numbers require rather than what management preferred to say.

What follows organizes five companies across the contracting dimensions that actually matter before you sign a 2026 completion program. Margins and leverage matter for obvious reasons. Pricing trajectory, customer concentration, capital posture, and management tone matter because they determine how this vendor behaves in a renewal negotiation and what happens to your program if their internal situation changes mid-contract.

Patterson-UTI (PTEN)

Q4 EBITDA margin: 19.2% — up 60bps sequentially on flat revenue of $1.15 billion. Margins expanding while revenue holds flat is a cost structure story, not a pricing power story. The company’s Q4 free cash flow came in at $259 million against a $167 million analyst consensus — a 55% beat. Approximately $15 million of that came from customer prepayments for 2026 work, which is itself a negotiating data point: counterparties who prepay are telling you something about their view of available alternatives. Net debt down to $822 million from $1.06 billion at Q3, at 0.9x net debt/EBITDA. No senior note maturities until 2028. Total liquidity $739 million. Fitch affirmed BBB- with stable outlook in August 2025. A 25% dividend increase to $0.10 per share — a signal of confidence in sustained cash generation that is harder to fake than a management quote.

The vendor that raises its dividend 25% in a flat market has decided it does not need your volume badly enough to discount for it. That is the first thing to understand about negotiating with PTEN in 2026.

What the pricing language actually says: Management described Q4 rates as “steady” — the same word used in Q3, which they also called “steady.” Barclays analyst Sungeun Kim identified the gap: Q1 2026 Completion Services gross profit guidance of $95 million represents a roughly 14% sequential decline against activity characterized as only “slightly” lower. Her question: “That would seem to imply a not insignificant pricing decline. Is that a fair assessment?” CEO Andy Hendricks responded “Not at all” and attributed the gap to weather and mix. Companies that are genuinely holding price explain the arithmetic. Companies managing the narrative change the subject. The gap between the word “steady” and the implied math in the Q1 guide is real and belongs in your model.

Completion services contract structure — what the frac business actually looks like commercially: PTEN’s completion programs reset annually. Dedicated fleet pricing is negotiated at year-end and “locked in place” for the coming year — management’s language, not a summary. Spot work fills calendar white space between dedicated programs, at whatever the market clears. Approximately 10% of completion work has shifted to performance-based agreements, where compensation is tied to efficiency outcomes rather than a fixed stage or hourly rate. That last number matters for a renewal negotiation: the buyer who enters a performance-based discussion expecting to benchmark against a simple dayrate is negotiating the wrong contract type against the wrong variable. The customer prepayment signal — roughly $15 million in advance payments for 2026 work embedded in the Q4 FCF beat — tells you something about how the customer base is sorting. Counterparties who prepay have decided that their alternative is worse than locking in early. Counterparties who did not are either testing the spot market or waiting to see if Q1 weather and pricing headwinds create an opening. Knowing which of those you are is the operational intelligence question.

Where the pricing power is real: PTEN was explicit that natural gas-capable equipment — roughly 85% of active fleet horsepower by year-end 2026 — is operating at near-full utilization with “minimal spare capacity.” Management predicted the gas fleet would be “essentially sold out” by Q2 2026. If that proves accurate, and Haynesville and Appalachian activity responds to LNG-driven demand as modeled, the pricing leverage dynamic for gas-capable units shifts sharply toward the vendor in H2 2026. The Permian is a different market. West Texas pricing is, per PTEN’s own words, “certainly still competitive” relative to gas basins. Buyers sourcing Permian oil-weighted programs have more room in 2026 than buyers competing for gas basin capacity.

What management conceded under pressure: When calendar white space appears, PTEN fills it at spot, and spot means a lower price. Exact language: “if there’s any white space in the calendar... we have to fill some dedicated work with some short-term spot work, maybe we take a little bit lower price to do that.” That admission is a procurement playbook. A buyer who delays award decisions or deliberately creates scheduling gaps is creating exactly the condition that triggers PTEN’s spot market behavior. Whether that tactic makes sense for your program depends on how much program continuity risk you are willing to carry in exchange for a rate concession.

Customer concentration: Top customer at 12% of consolidated revenue for full year 2025. Top five: 39%. Top ten: 57%. Roughly 60% of revenue from the 15 most active U.S. E&Ps — large public and large private, structurally resilient but also the customers PTEN cannot afford to lose, which provides leverage to that cohort in renewal discussions. Performance-based agreements now cover approximately 10% of completion work — a contract structure that protects PTEN’s revenue from dayrate deflation by tying compensation to efficiency outcomes. The buyer who pushes purely on rate against a performance contract is negotiating the wrong variable.

Halliburton (HAL)

Q4 adjusted operating margin: 19.8% — the highest in the peer group, up 130bps sequentially. Before using this as a benchmark for anything, read the Q&A. CFO Eric Carre confirmed that over half of the guided 300bps margin compression for Q1 2026 is explained by the non-repeat of year-end completion tool sales — a transactional event that inflates Q4 and disappears from Q1. Anyone using HAL’s Q4 C&P margin as a run-rate for North American frac profitability is using the wrong input. The underlying run-rate is closer to the Q1 guide than the Q4 print.

Balance sheet: Net debt $4.95 billion at year-end, down from $5.4 billion in Q3. Net debt/EBITDA 1.2x at investment-grade (BBB+/A3). Total liquidity approximately $5.7 billion. Q4 FCF $875 million; full year $1.9 billion. The company returned 85% of annual FCF to shareholders and repurchased $1 billion in stock during 2025. The 2026 CAPEX budget is down 30% from 2025 — the largest percentage reduction in the peer group.

The contracting implication of the North America guidance: HAL explicitly guided to a high-single-digit decline in North American revenue for 2026. The highest-margin vendor in this peer group is deliberately shrinking its North American completions business. CEO Jeff Miller: “Pricing reaches a point where it’s not... companies aren’t investing in it. We are moving equipment away from it.” That statement functions as a credible commitment to supply withdrawal only because HAL actually has an alternative — 59% of total revenue is international, growing 7% sequentially in Q4. A vendor who genuinely does not need North American volume to cover overhead can execute a withdrawal threat. Most vendors cannot. HAL can. This changes the negotiating dynamic in a specific way: HAL is not a vendor who will chase your program at a concession to maintain utilization. The question for E&P procurement is whether the premium over comparable conventional services is justified by the performance differential — because HAL is not offering a discount path to find out.

Customer and counterparty risk — one number that belongs in any complete HAL analysis: The company entered into $750 million in Credit Default Swaps to hedge receivables exposure to Pemex, which represents 11% of total receivables. This reflects a working capital risk in the consolidated entity that does not appear in the North American completions P&L but affects overall financial positioning. The significance for E&P buyers is indirect but real: a company managing concentrated international credit risk may price NAM work differently than its standalone NAM margins would suggest.

What the stacking strategy means operationally: HAL stacked fleets in Q3 and continued stacking in Q4 — voluntarily, at rates that management characterized as “not economic.” The crews associated with stacked fleets do not stay on payroll indefinitely. The personnel who run HAL’s most complex completion systems are specialized and mobile. A sustained period of voluntary capacity withdrawal has implications for the speed and cost of reactivation if demand recovers sharply. The vendor who confidently withdraws supply in a soft market is also the vendor who faces the steepest reactivation cost if the market inflects faster than they modeled. That asymmetry is worth tracking.

Liberty Energy (LBRT)

Q4 EBITDA margin: 15.2%, up from Q3’s 13.5%. Revenue grew 10% sequentially — the strongest sequential revenue print in the peer group. Balance sheet: net debt $219 million at year-end at approximately 0.4x net debt/EBITDA — the lowest leverage ratio in the group. Total liquidity $281 million, up from $146 million at Q3.

The pricing disclosure that should be in your file: Liberty was more specific than any other company in this peer group, and the specificity is analytically useful precisely because the content is unfavorable to the vendor. CEO Ron Gusek quantified the Q4 RFP season impact as “low to mid-single digits” below H2 2025 rates — confirming that dedicated customers who demonstrably value Liberty’s operational record still extracted price concessions during the annual contract reset. His description of the trend: “a slow and steady trend downwards” from the mid-2022 peak. Goldman Sachs analyst Ati Modak pushed directly on the contradiction between “flight to quality” and concurrent pricing concessions. Gusek’s answer: competitors “find white space” and defend share with price, “we are not immune to these market forces.” That sentence is worth its own line item in any Liberty renewal discussion. The vendor who publicly acknowledges they cannot hold price against commodity competitors is giving you the benchmark for your next negotiation.

Why this matters for contracting beyond the rate: The RFP process that drove these concessions ran during Q4 2025, when activity was soft and buyer leverage was highest. The rates set in that process are locked for 2026. If the gas demand inflection that LBRT (and PTEN) are predicting for H2 2026 materializes, those rates will look increasingly favorable to Liberty — and they will have the utilization data and market conditions to push for higher rates at the next annual reset. Operators who locked term in Q4 2025 captured the trough pricing. Operators renewing in Q4 2026 may be negotiating in a different market. The timing of your commitment is itself a pricing decision.

Capital allocation — the most consequential signal in this quarter’s disclosure: 2026 completions CAPEX approximately $250 million, of which roughly $175 million is maintenance. Power generation CAPEX: $725 million to $900 million, targeting 3 gigawatts of deployed capacity by 2029. That ratio — $3.50 to $3.60 of power capital for every dollar of frac capital — is a clear statement of where growth investment is going. CFO Michael Stock confirmed capex is “markedly shifting” toward power with completions in maintenance mode. The specific phrase: “No digi builds in 2026.”

What this means for a multi-year completion contract: Liberty’s frac business has formally entered harvest mode to fund a power infrastructure buildout. The business runs well today — the Q4 operational numbers support that. The question for an E&P operations team evaluating a 2026-to-2028 term contract is whether frac continues to receive the management attention and equipment investment that the current operational record reflects. The $800 million per year flowing to LPI is not flowing to frac fleet quality. Maintenance CAPEX budgets on existing equipment set the reliability trajectory. The gap between “maintenance mode” and “adequate maintenance” is what your completion cost per foot eventually measures.

Customer concentration — monitor the ExxonMobil exposure: Occidental Petroleum and ExxonMobil each exceeded 10% of 2025 revenue. The ExxonMobil position warrants specific attention given ProPetro’s parallel and disclosed non-renewal situation for two FORCE electric fleets contracted to the same operator. Whether the ExxonMobil programs in Liberty’s portfolio carry equivalent renewal risk is not publicly disclosed — but a procurement team that knows ExxonMobil is actively restructuring its vendor commitments across multiple service providers should be asking the question directly rather than waiting for it to appear in a 10-K.

ProPetro (PUMP)

Q4 EBITDA margin: Estimated at approximately $32 to $34 million total — a run rate suggesting roughly 11-12% margin on the completion services business. The organizing fact for ProPetro in 2026 is not the margin. It is the customer concentration.

The ExxonMobil situation in plain arithmetic: ExxonMobil accounts for approximately 19.7% of total ProPetro revenue. Per ProPetro’s own prospectus filing dated January 2026, the agreement governing two FORCE electric fleets for ExxonMobil expires in late 2026 and is “not expected to be renewed.” Permian Resources: 14.9% of revenue. EOG Resources: 10.6%. Top five customers: 58.8% of revenue. Top ten: 75.3%. ProPetro derives 98.5% of fleet revenue from the Permian Basin. When approximately one-fifth of your revenue from a single customer is disclosed as non-renewing in your own SEC filing, the follow-on question for every other customer in the portfolio is whether their program pricing reflects a vendor under competitive pressure to replace that volume. It usually does.

The equity raise tells the story that the earnings call could not: ProPetro launched a 12.5 million share equity offering in January 2026 specifically to fund ProPWR growth capital. The timing — equity raise at cyclical lows, simultaneously with a disclosed anchor-customer non-renewal — reflects a vendor whose organic cash flow from the core completion business is insufficient to fund its strategic pivot. Management has stated the completion business is in “maintenance mode.” The ProPWR book includes 220 MW of contracted power, including a 10-year Permian operator deal and a data center contract. The strategic logic is sound. The contracting implication for E&P buyers is more immediate: a vendor simultaneously managing a 20% revenue hole from a non-renewing anchor customer and redirecting capital to an adjacent business is a vendor whose internal resource allocation decisions deserve explicit conversation before any program is committed. The best equipment and most experienced crews go to the anchor programs. Understanding where your program ranks in that allocation is not a theoretical question in ProPetro’s current situation — it is the operational risk question.

The Permian density is real and belongs in the analysis: ProPetro’s operational density in the Permian Basin — 98.5% of revenue, infrastructure concentrated around a single geography — creates genuine logistical advantages that distributed competitors cannot easily replicate. Mobilization costs, crew familiarity, and local supply chain depth all improve at this concentration level. For a buyer running a Permian-only program with no geographic flexibility in their vendor set, ProPetro’s basin expertise offsets some of the customer concentration risk. The relevant question is what rate premium that density justifies relative to the negotiating leverage available from the current financial situation.

What “disciplined” actually means in this context: Management characterized Q4 pricing as “disciplined” — the word service companies use when they are trying to hold a line they find harder to hold than the word implies. Q3 pricing language was “weak.” “Disciplined” is the upgrade. The arithmetic behind the upgrade is that ProPetro reduced active fleet count from 13-14 in Q2 to 10-11 in Q3-Q4 — deliberately idling fleets to avoid working at sub-economic rates. A vendor who idles 20-25% of their fleet to protect pricing has made a specific judgment about where the pricing floor is. It also means 20-25% of that vendor’s revenue-generating capacity is sitting idle, which changes the math on how aggressively they need to fill what remains at what rate.

ProFrac (ACDC)

Start with the credit event, because everything else is organized around it: S&P Global Ratings lowered ProFrac’s issuer credit rating to CCC in January 2026 before withdrawing it at the issuer’s request. The withdrawal does not reverse the downgrade or the underlying conditions that produced it. Net debt: approximately $1.04 billion against $58 million in cash at Q3, at roughly 3.4x net debt/EBITDA — the highest leverage ratio in this peer group by a factor of approximately three. Q3 free cash flow: negative $29 million. The company executed a fourth amendment to its credit agreement in December 2025, reducing amortization payments and deferring leverage ratio testing to March 2028. The ABL facility becomes current in early 2026.

What 3.4x leverage at CCC-equivalent capital costs means operationally: The interest expense on $1.04 billion in net debt at current borrowing costs is consuming operating cash flow at a rate that limits the capital available for equipment maintenance, crew quality, and program investment. The $100 million annualized cost-saving initiative — covering SG&A (reduced 17% sequentially in Q3) and a headcount reduction executed October 2025 — is the organizational response to that constraint. Cost-cutting preserves liquidity. It does not improve the equipment maintenance posture. Q3 Stimulation Services segment EBITDA margin: 6%. West Texas pricing described as “competitive” — the word in this industry that means the vendor is not setting the clearing price, the market is.

The cost structure advantage is real and belongs in this analysis: ProFrac’s in-house manufacturing produces frac equipment at approximately $540 per HHP versus an industry average of approximately $861 per HHP — a 37% cost advantage on newbuild equipment. The company also owns sand mines and chemical manufacturing, capturing margin at multiple points in the completion cost chain that non-integrated competitors pay through. This is not trivial. The provider who can build at 63 cents on the dollar has a structural cost floor that competitors without equivalent integration cannot match. If ProFrac’s financial structure were different, this cost basis would translate into durable pricing power on the low end of the market.

The problem the cost advantage cannot solve: The interest expense on $1.04 billion in net debt at CCC-equivalent capital costs is consuming that operating advantage faster than the business regenerates it. The fourth credit agreement amendment is a deferral of the leverage ratio test, not a resolution of the underlying leverage. The buyer who negotiates the deepest rate concession from ProFrac — and the leverage for that negotiation is genuinely high, because the vendor needs cash flow to service debt and will price work accordingly — is the buyer with the highest exposure if a restructuring process interrupts mid-contract. Those two facts describe the same entity observed from opposite ends of the same problem.

The commercial pivot and its execution risk: Management’s 2026 strategy is to move from spot-heavy to dedicated programs, targeting 90%-plus of active fleet capacity on committed work. The rationale is sound — spot market pricing has converged with dedicated pricing, eliminating the premium that previously made spot market exposure attractive, and the “head fakes” in Q3 spot activity caused material operational inefficiency. A dedicated program base provides cash flow stability that the debt structure requires. The execution risk: restructuring commercial terms from a position of financial distress limits how aggressively ProFrac can negotiate the terms of those dedicated agreements. A counterparty who knows the vendor needs the volume to service its debt is a counterparty with a structural negotiating advantage in every discussion about price, equipment quality, and crew assignment.

The specific data point most buyers overlook: Spot pricing and dedicated pricing have converged. Per management: “spot and term pricing [are] pretty in line relative to each other.” The historical premium for accepting spot market volatility has disappeared. This means the buyer who previously used spot market pricing as a lever in dedicated contract negotiations has lost that benchmark. The market has done the compression for them. What remains is leverage derived from the vendor’s financial position — and that leverage is substantial and real, but it comes with the program continuity risk described above.

The Vendor Scorecard

Five observations the matrix reveals that the individual dossiers do not.

1. HAL and PTEN score identically in total, but for structurally different reasons. HAL’s score is anchored by balance sheet strength and willingness to credibly withdraw supply — genuine leverage backed by an international revenue base that does not require North American volume. PTEN’s score reflects operational consistency and margin quality. Where they diverge is on frac management attention: HAL’s active capital investment is migrating toward Zeus, which is a different negotiation than conventional tier. PTEN still has conventional capacity, even if it is being managed at the margin.

2. LBRT’s composite score (23) understates the operational quality of the frac business and overstates the risk — the score reflects the capital allocation signal, not a current performance problem. The business runs well today. The question is the 2026 and 2027 forward-looking answer when $800 million per year in growth capital is flowing to LPI rather than frac. That uncertainty loads the “management attention” and “program continuity” dimensions.

3. PUMP’s score of 18 reflects a specific, identifiable event rather than a general posture of weakness: the ExxonMobil non-renewal. A vendor managing a 19.7% revenue hole while raising equity to fund an adjacent pivot is a vendor whose negotiating position has narrowed for a specific window. The basin depth and operational density in the Permian are genuine, and they matter for a Permian program.

4. ACDC scores higher on buyer pricing leverage (5) than any other vendor and lower on program continuity (1) than any other vendor. Those two numbers describe the same entity observed from opposite ends of the same problem. The operator who negotiates the deepest rate concession from ProFrac is the operator with the highest exposure if a restructuring process interrupts mid-contract. That trade-off does not resolve without a specific view on whether the credit trajectory stabilizes.

5. The peer group’s aggregate management attention score on frac is the number that deserves the most scrutiny from an E&P operations team doing multi-year planning. Three of the five vendors in this peer group are actively redirecting growth capital away from frac. The equipment in the commodity tier is the equipment that exists today, managed on maintenance budgets that are at or below the floor required for standard PM intervals. That is the market you are contracting into in 2026.

TLDR: The Tool Manufacturer Won

Remember the tool manufacturer from the beginning of this piece. They had an old tool, a new tool, a Sherman Act problem, and a CFO with a face. The Q4 2025 earnings cycle is the closing chapter of their story: the old tool is gone. Not stacked. Not temporarily idled. Gone. HAL pulled the last of the conventional excess capacity off the market and is holding it in reserve to be reactivated at a price of their choosing, which is a different thing than capacity that is available at any price. When the seller controls the release valve, the buyer does not control the pressure.

The cascade from here is straightforward if you follow it in sequence. Conventional diesel is off the table as a competitive tier. Tier 4 DGB is now what diesel used to be — the commodity floor, the equipment receiving maintenance-only capital, the fleet that every service company still runs but none of them are growing. Next-generation electric and mechanical drive is where newbuild capital is going, and most of that capital is chasing power infrastructure and international markets, not more frac fleets. The service companies have spent four years methodically eliminating their own worst alternative — cheap conventional supply they could not profitably stack — while simultaneously building yours. That is the complete picture.

Here is what it means in practice.

If you are running DGB equipment in 2026, stop optimizing for rate and start optimizing for reliability. The capital that used to maintain your vendor’s fleet is leaving. Simul-frac customers get the maintenance CAPEX that remains, because simul-frac is where the margin is. A further concession on stage rate from a vendor whose fleet is being systematically underfunded does not reduce your completion cost. It increases it — it just moves the cost from the invoice to the NPT report. Build performance standards and equipment quality provisions into the contract before you negotiate price. The vendor running a properly funded maintenance program signs those terms without objection.

Start with a should-cost model. You do not need Kalibr Pulse to build one. Public financial disclosures give you fuel consumption benchmarks, maintenance cost ranges, crew cost data, and the margin math at current activity levels. What you are looking for is whether the bid in front of you is fundable at the performance standard your program requires, or whether the vendor is borrowing from future reliability to win today’s award. The gap between a sustainable margin and the clearing price in your RFP is the leading indicator of future downtime. Build the model. Read the gap.

Long-term, the question is optionality. The technology choices you are making today in development planning have a longer shadow than they appear to. A Zeus-optimized program is a program designed around pad configurations, water logistics, and well sequencing that is incompatible with conventional fleet economics. That is a feature when Zeus is available and performing. It is a constraint when it is not. Before you build pad architecture around any single provider’s fleet specifications, understand what you are giving up and for how long.

If you have a consistent completion program outside the Permian and you are not currently in conversations with ProPetro, have the conversation. The Permian service market is facing structural headwinds — too many fleets concentrated in one basin with anchor customer commitments unwinding. ProPetro has genuine operational density in the basin that does not travel easily, but the financial math of their current situation creates negotiating conditions that will not persist indefinitely. A non-Permian E&P with a reliable program is exactly the customer they need right now, which means you have leverage you will not have in eighteen months.

Finally: if this situation looks familiar, it should. What happened in the frac market over 2023 to 2025 — systematic retirement of commodity capacity, capital concentration in premium equipment, service company leverage rebuilding as the commodity alternative disappeared — is structurally identical to what happened in compression in the early 2020s. The compression playbook that worked then works here. Think about what you did differently when you understood that the old commodity compressor fleet was not coming back, and apply the same logic to what you are signing in frac today.

The tool manufacturer built three new tools, retired the old one, and handed you a market where the commodity option no longer exists. The question they are waiting to see answered is whether you noticed before you signed the renewal.

Subscribe now