Behind the Meter (BTM): Why Oil & Gas Companies Are Also Becoming Power Companies

Energy News Beat Channel — September 11, 2026

Hector Acevedo put the industry’s new map on the table this week: AI did not just raise electricity demand. It collapsed the old division of labor. Utilities no longer have a monopoly on electrons. Oil and gas companies no longer stop at molecules. Hyperscalers no longer wait in interconnection queues. They buy fuel, generation, cooling, and uptime as one system.

That is the behind-the-meter thesis. Power is generated on the customer’s side of the utility meter and delivered under a private contract. No multi-year transmission study. No socialized grid tariff stack. The commercial relationship is steel, fuel, and a take-or-pay offtake. For companies that already own gas, pipe, turbines, field logistics, and modular fabrication, that is not a side quest. It is a second operating model sitting on assets they already know how to run.

Acevedo’s scorecard is the right starting point: Chevron’s Energy Forge One and Project Kilby with Microsoft; Williams’ Socrates complex next to Meta in Ohio; SLB, Halliburton, and Baker Hughes moving from oilfield services into power, thermal management, and modular data-center infrastructure. Add two names that now sit at the center of the same trade: EQT, which is monetizing Appalachian gas into power-linked contracts without building the plants itself, and Liberty Energy, which is turning a completions company into a power-as-a-service platform.

The demand shock is no longer theoretical. Goldman Sachs has U.S. data-center load rising from about 31 GW in 2025 to 66 GW by 2027. McKinsey has put the 2030 figure near 121 GW. Acevedo cited roughly 19 GW of U.S. gas generation already associated with Microsoft, Amazon, Meta, and Google across announced, contracted, permitted, or developing projects. Grid queues still measure in years. That gap is the market.

The financial question for Energy News Beat readers is narrower and harder: has this pivot already helped cash flow and shareholder returns, or is it still a story about 2028?

The honest answer from second-quarter 2026 earnings is mixed by design. The companies furthest into contracted power — Williams, Baker Hughes, SLB — are already booking orders, raising guidance, or printing a new revenue line. The companies building multi-gigawatt campuses — Chevron and Liberty — are spending now for mid-teens, take-or-pay cash flows later. EQT is collecting a different prize: premium gas pricing tied to power markets with almost no incremental generation capex.

The architecture: why BTM fits oil and gas.

A BTM plant and its load are one electrical system. The generator and the customer design, permit, and energize together. Front-of-the-meter megawatts have to be studied against every other user on the shared grid. Behind the meter, the only parties in the room are fuel, steel, and the offtaker.

That is why the oilfield skill set transfers. Pressure pumpers already run large generator fleets. Midstream companies already sit on the gas molecule and the right-of-way. Service companies already fabricate modular kits, manage remote high-load operations, and sell aftermarket uptime. Hyperscalers need firm power on a 24- to 36-month clock, not a 2032 interconnection date. Gas-fired generation, batteries for ride-through, and modular cooling are the stack that can hit that clock.

The cash-flow logic is just as important as the engineering. A 20-year take-or-pay PPA with a AAA-rated hyperscaler is not a Henry Hub strip. It is a utility-like contract sitting inside an oil and gas balance sheet. That is the point Chevron, Williams, and Liberty keep making to investors: uncorrelated duration against a commodity cycle that still sets the multiple on the legacy business.

Chevron: the multi-gigawatt template

Project Kilby is the cleanest illustration of Acevedo’s argument. Chevron’s Energy Forge One signed a 20-year take-or-pay PPA with Microsoft for 2.67 GW of firm behind-the-meter capacity at a co-located West Texas data-center complex. Management has called it the only multi-gigawatt BTM project with a long-term PPA at that scale. FID is targeted for later in 2026. First power is aimed at 2028. Returns are described as mid-teens. Cash flows are designed to be independent of oil and gas prices.

Kilby did not make Chevron’s second-quarter print. What made the print was the legacy machine: $12.1 billion in reported earnings, $12.0 billion adjusted, 21 percent return on capital employed, $22.6 billion of operating cash flow ($19.7 billion excluding working capital), and record U.S. production. Management still put the Microsoft contract in the headline of the release. That is not decoration. It is a signal that Chevron wants the market to underwrite a second cash-flow stream that does not live and die with Brent.

The investor implication is duration, not this quarter’s free cash flow. Kilby is inside the existing $18–19 billion capital framework. If Chevron can repeat the model in the Midwest, Rockies, and Gulf Coast — management has said it is already in those conversations — the company is building a contracted power annuity on top of Permian gas that would otherwise fight Waha blowouts. Localized gas-to-power is also a structural bid under associated gas. That matters for basin cash flow even before the first electron is sold.

Williams: first power, first guidance lift

Williams is the furthest along among Acevedo’s midstream names, and the second-quarter numbers show it.

Socrates Phase 1 in New Albany, Ohio — built for Meta’s campus — went into service in late July 2026, inside 18 months of commercialization, with about 200 MW of first power and a ramp to full phase capacity. Socrates North and South together are expected at 556 MW, backed by a 10-year, primarily fixed-price PPA. The broader Socrates program in public filings has grown into a multi-facility, multi-gigawatt gas-plus-battery complex. Additional Power Innovation projects — Neo, Aquila, Socrates the Younger, Apollo — extend the book into 2027–28.

Q2 2026 results: GAAP net income of $827 million, or $0.68 per share, up 51 percent year over year; adjusted net income of $614 million; adjusted EBITDA of $1.921 billion, up 6 percent; cash flow from operations of $1.376 billion; available funds from operations of $1.450 billion, up 10 percent; dividend coverage of 2.26x on an AFFO basis. The company raised 2026 adjusted EBITDA guidance by $200 million at the midpoint, to $8.4 billion. Some of that lift is the $5.5 billion Momentum Midstream acquisition. Some of it is the power platform starting to show up in the run-rate.

The financing structure matters as much as EBITDA. Williams closed a $5.34 billion Power Innovation joint venture with Blackstone, including $4.4 billion to fund 49 percent of growth capex across five projects at a capped 6.4 percent cost of equity. That is how a midstream company builds a generation fleet without blowing the 3.5x–4.0x leverage box. Earlier this year, management said the power platform could deliver about $1.4 billion of annual contracted EBITDA by 2029. If that holds, BTM is not a narrative overlay. It is a visible share of the 11 percent-plus long-term EBITDA growth target Williams just raised.

For shareholders, the near-term return mechanism is still the dividend machine: $0.525 per share in June, coverage above 2x, and a growth capex budget of $7.3–$7.9 billion that now includes power as a core line, not a pilot. The stock is being asked to re-rate from “gas pipeline” toward “contracted energy infrastructure.” Socrates in service is the first hard evidence that a re-rate can be earned rather than advertised.

EQT: selling the molecule into the power stack

EQT is not building Kilby. It is doing something more capital-light and, in the near term, more cash-flow immediate.

In the second quarter, the company signed a 10-year definitive agreement to supply 325,000 Dth/d of gas to Competitive Power Ventures’ planned 2 GW Shay Energy Center in Doddridge County, West Virginia. Pricing is linked to PJM power prices, not a local gas index. Management called it the second West Virginia combined-cycle project EQT has helped catalyze after Wolf Summit, and the second contract with this power-linked structure. At the forward strip, the deal is expected to clear a material premium to in-basin pricing, with no EQT capex in the plant.

That is a producer version of the utility framework: long-duration offtake, priced off the constrained power market that data centers are tightening, using gas EQT already plans to produce. The rest of the commercial book points the same direction — accelerating MVP Southgate, a 0.5 mtpa LNG offtake from 2028 expected to add about $45 million of 2028 free cash flow at recent strips, and a marketing strategy built around becoming the “partner of choice” for power generators and data-center developers in Appalachia.

Q2 2026 operations supported the cash: 634 Bcfe of sales, above the high end of guidance; capex of $666 million, 9 percent below the low end; net cash from operations of $1.048 billion; adjusted EBITDA attributable to EQT of $1.067 billion; free cash flow attributable to EQT of $330 million. Full-year production guidance went up by about 90 Bcfe. Capital spending guidance came down $25 million even after pulling $85 million of Southgate contributions forward.

EQT’s power pivot is not yet a standalone P&L segment. It is already a basis and realization story. Every Mcf that leaves the basin on a PJM-linked contract instead of a discounted local index drops straight through to free cash flow. That is how a gas producer joins the hyperscaler stack without becoming a regulated utility.

Liberty Energy: the completions company that is buying turbines

If Chevron is the integrated-major template and Williams is the midstream template, Liberty is the oilfield-services template taken all the way to ownership.

Liberty Power Innovations is no longer a slide. In January, the company signed with Vantage Data Centers for up to 1 GW over five years, including 400 MW reserved for 2027. In May it awarded Bergen Engines more than 500 MW of on-site gas generation. In July it formed a joint venture with PowerBridge for a planned 2 GW West Texas campus — first phase more than 300 MW, first power targeted for late 2027 — and a strategic alliance with SLB for modular BTM infrastructure. Management now cites visibility to about 3 GW of generation capacity through 2029 and early 2030. Total capital for that pipeline is estimated at $5–6 billion. Return targets: 5- to 6-year cash-on-cash payback and 17–18 percent unlevered returns.

Q2 2026 still looks like a completions company in transition. Revenue was $1.2 billion, up 14 percent year over year and 16 percent sequentially. Net income was $43 million, or $0.26 per share. Adjusted EBITDA was $151 million, down from $181 million a year earlier but up from $126 million in Q1. The company paid a $0.09 dividend. Cash ended the quarter at $555 million with about $1.0 billion of liquidity, aided by new convertible notes. 2026 capex guidance is about $1.5 billion, heavily weighted to long-lead power equipment. Q2 net capex and deposits included $71 million already earmarked for power.

This is the cash-flow paradox of the whole theme. Liberty’s power business is not yet carrying the P&L. It is absorbing capital. The completions franchise is still funding the dividend and the story. Investors have already voted with the multiple at times — the stock jumped to a record earlier this year on the 3 GW target — then sold off 13.7 percent on the Q2 print when the capex bill came into focus. That is what a framework shift looks like in real time: the market pays for utility-like duration, then demands proof that 17 percent unlevered returns survive turbine lead times, air permits, and first-of-kind site work.

If Liberty hits the 2027–29 COD dates, the cash-flow mix changes. Power-as-a-service is contracted, owned, and operated. That is closer to a midstream MLP or an IPP than to a frac fleet. Until then, shareholder returns depend on the old business staying utilized enough to carry the new one.SLB: data centers are already in the income statement

SLB is the Acevedo name where the pivot is no longer prospective.

Data Center Solutions revenue rose 33 percent sequentially and 80 percent year over year in Q2. First-half 2026 revenue in the business was up 63 percent. Management said the unit is on track to exceed a $1 billion annualized run rate by year-end 2026 and more than $2 billion as it exits 2027. SLB has been selected as a delivery partner on Meta’s planned 1 GW Sturgeon County, Alberta campus. The company says it has shipped more than 1.3 GW of prefabricated modular infrastructure since April 2024 and expects to exceed 2 GW of deliveries globally this year. The Liberty alliance puts SLB modules and project execution next to Liberty’s BTM generation. The $3.4 billion Kelvion acquisition — about $4.1 billion including assumed debt — buys data-center cooling and thermal management at scale. Combined, SLB and Kelvion’s data-center businesses are expected to generate more than $2 billion of revenue and about $300 million of adjusted EBITDA on a 2026 pro forma basis.

The core company still pays the bills. Q2 revenue was $8.97 billion. Adjusted EPS was $0.55. Adjusted EBITDA was $1.90 billion. Cash from operations was $1.36 billion. Free cash flow was $716 million. Those figures are not from a data-center company. They are a diversified energy-technology company whose fastest-growing line item is now the hyperscaler campus. That is how the framework helps investors today: it offsets softness in traditional international product lines and gives the equity a growth sleeve the oilfield cycle does not control.

Baker Hughes: orders first, cash later — and cash is already showing up

Baker Hughes has gone further than almost anyone in turning power and industrial infrastructure into the majority of the company. Management has said roughly 60 percent of revenue now comes from infrastructure and industrial markets rather than traditional upstream services. The Chart Industries acquisition is the thermal and industrial-gas half of that bet.

Q2 was an orders quarter. Total orders were $10.5 billion. IET orders were a record $7.1 billion, doubling year over year. Remaining performance obligations reached $40.1 billion, including record IET RPO of $37.1 billion. Power Systems booked $2.6 billion of orders and 2.7 GW of generation, including a 1.3 GW NovaLT award from Dynamis for mobile data-center and oilfield power and a Kodiak Gas Services framework starting at 1 GW and running up to 1.8 GW. IET has secured $4.2 billion of data-center-related orders since 2025, $3.2 billion of that in the first half of 2026. The chart has booked about $600 million of data-center orders over two years.

The cash line is the part investors should not skip. Adjusted EBITDA was $1.231 billion. Free cash flow was $1.109 billion in the quarter. Full-year free-cash-flow conversion is still guided to 40–45 percent even after Chart-related costs. That is the opposite of Liberty’s current mix: Baker Hughes is selling equipment and services into other people’s BTM buildout, so the power boom hits orders and cash before it hits a 20-year owned-plant depreciation schedule. The Chart stub-period guide — $1.85–$2.25 billion of revenue and $300–$400 million of EBITDA through year-end — is the cooling-and-cryogenics layer Acevedo flagged when he said the battleground has moved past generation into thermal management.

Halliburton and VoltaGrid: the distributed-generation option

Halliburton’s path is partnership capital rather than a 2.67 GW owned campus. The company and Blackstone Tactical Opportunities put $1 billion into VoltaGrid — $775 million primary, $225 million secondary — to expand BTM generation for data centers, microgrids, and industrial load. VoltaGrid has cited a 7.5 GW order book through 2030 and modular QPac nodes of up to 20 MW, scalable to about 200 MW under a minor-source air permit. Halliburton and VoltaGrid had already targeted distributed power for data centers, including 400 MW of modular gas systems for the eastern hemisphere.

Halliburton’s own Q2 still reads as a completions-and-drilling company: $5.7 billion of revenue, $534 million of net income ($0.64 per share; $0.55 adjusted), $824 million of operating cash flow, $668 million of free cash flow, and about $200 million of buybacks. Power is not yet a reported segment. The return mechanism, for now, is capital discipline in the core plus an equity-style claim on a distributed-generation platform that already has a multi-gigawatt book. That is a cheaper way to buy optionality than Liberty’s $5–6 billion owned fleet — and a slower way to change the cash-flow mix.

What the numbers actually say about cash flow and returns

Strip out the slogans, and Q2 2026 leaves a four-part scorecard.1. Contracted power is already improving cash-flow quality where it is in service or in the order book. Williams has first electrons, a Blackstone funding partner, higher AFFO, and a raised EBITDA guide. Baker Hughes is converting data-center orders into record backlog and a $1.1 billion free-cash-flow quarter. SLB is running a data-center line that is compounding at 60–80 percent and is large enough to matter at the group level. Those are not 2030 stories.2. Owned generation is still a capex story. Chevron’s Kilby and Liberty’s 3 GW pipeline will not drop meaningful operating cash until 2027–29. Liberty is explicit: $1.5 billion of 2026 capex, $5–6 billion through the decade, paybacks measured in years. That can raise the equity multiple if investors trust the 17 percent unlevered math. It can also suppress near-term free cash flow and punish the stock on any slip in COD.3.

Producers can capture the framework without owning the plant. EQT’s PJM-linked CPV contract and power-demand marketing are a realization upgrade. Free cash flow attributable to EQT of $330 million in a single quarter, with rising volume guidance and falling spend guidance, is the producer version of “becoming a power company.” The power plant is on someone else’s balance sheet. The premium is on EQT’s.4. Shareholder returns are splitting along business model. Halliburton and SLB are still returning cash from the oilfield machine — buybacks and free cash flow — while they bolt on infrastructure. Williams is covering a growing dividend with AFFO that now includes power. Liberty is paying a modest dividend while levering the balance sheet for turbines. Chevron can fund Kilby inside a program that already produced a 21 percent ROCE quarter. The framework helps most where it either (a) is already contracted and in service or (b) is sold as equipment and modules rather than owned megawatts.

None of this is risk-free. Turbine backlogs are long. Air permits and community opposition can slip a COD by years. Take-or-pay contracts only help if the plant gets built and the offtaker’s campus shows up. ERCOT, PJM, and state commissions are writing new rules for large private generation as fast as developers can pour pads. A BTM campus that later wants a grid tie still collides with the queue it was designed to avoid. And a frac-driven cash engine that funds a power buildout can stall if North American completions roll over before first power.

Acevedo’s conclusion still holds. AI is not only demand for more electricity. It is a new industrial architecture. Technology companies become energy buyers. Oil and gas companies become power providers. Oilfield service companies become infrastructure platforms. The winners will integrate fuel, generation, cooling, controls, and compute while keeping commercial and regulatory accountability in one stack.

The second-quarter tape says the architecture is no longer a white paper. It is in Williams’ in-service date, Baker Hughes’ order book, SLB’s run-rate guide, EQT’s power-linked gas contract, Chevron’s Microsoft PPA, and Liberty’s turbine deposits. Cash flow has started to follow the contracts that are live. Returns to investors will follow the contracts that stay on schedule.