The New Gusher: How Big Oil Became Big Power in the Age of AI

The Digital Oilfield · Field Note

Chevron, ExxonMobil and a wave of gas turbines are turning the American oil patch into the power supply for artificial intelligence — and rewriting what an energy company is.

By Prasad Selvaraj · September 8, 2026

The turbines arrived in West Texas the way the drilling rigs once did: on flatbed trailers, in convoys, trailing dust across a landscape that has spent a century learning to absorb the machinery of energy booms. But the crews bolting them into place near Midland were not chasing oil. They were building an electricity plant so large, and so single-minded in its purpose, that it has no customers at all except a server farm — and no real precedent in the hundred-odd years since the Permian Basin was first punctured for crude.

This is Project Kilby, a 2.67-gigawatt gas-fired power complex that Chevron and Microsoft agreed this summer to build under a 20-year, roughly $7 billion contract. It will not send a single electron to the public grid. Its entire output has been pre-sold to one buyer, for one purpose: keeping the lights on, and the chips cool, inside the data centers training the next generation of artificial intelligence.

2.67 GW

The capacity of Project Kilby — enough to power roughly two million homes — sold entirely to a single AI data-center customer under a 20-year, $7 billion contract.

Ten months ago, this would have been described as an oil company's side project. It no longer is. Across the industry this publication has spent two years chronicling — the sensors, the satellites, the software that turned oilfields into data streams — the biggest story of 2026 is not a well. It is a wire. Big Oil is becoming Big Power, and it is doing so at a speed that has caught even veteran energy analysts off guard.

The New Land Rush

Chevron's arrangement with Microsoft is only the most visible of a wave. ExxonMobil, which spent a century defining itself by what it pulled out of the ground, is now marketing a 1.5-gigawatt natural gas plant to data-center developers before it has even disclosed where the plant will sit — a project the company says will pair with carbon-capture technology to remove more than 90 percent of its emissions. In a separate venture with NextEra Energy and Google, Exxon is advancing a 1.2-gigawatt facility built to run alongside a hyperscale computing campus. Chevron, for its part, has already ordered seven GE Vernova turbines for delivery this year and is scouting sites across the South, the Western Interior and the Midwest for a fleet of gigawatt-scale plants, part of a broader ambition to control 3 gigawatts of dedicated data-center power within two years.

The Gigawatt Race

Dedicated power capacity committed to AI data centers by the two largest U.S. oil majors

Source: Chevron; ExxonMobil; Data Center Dynamics; BIC Magazine, 2026 disclosures.

“The customer interest is high,” Daniel Droog, Chevron's vice president of power solutions, has told industry audiences this year. “We're really focused on scale, speed and reliability.” It is a sentence that could have been lifted from any oilfield boom of the last century, aimed now at a customer that does not exist in the oil business at all: the hyperscaler, racing to plug in racks of processors faster than the public grid can plug in anything.

That is the mechanism driving all of this. Data centers built for artificial intelligence now draw fifty times the power of a conventional facility, and global data-center electricity demand is on pace to triple within three years. Utilities, bound by regulatory approval timelines and interconnection queues that routinely stretch past three years, cannot move fast enough. Oil and gas companies — who already hold the drilling rights, the pipeline capacity, the turbine contracts and decades of experience building large, remote, single-purpose infrastructure at speed — can.

Data Centers' Growing Share of U.S. Electricity

Share of total U.S. electricity consumption drawn by data centers

Source: U.S. data-center electricity consumption estimates cited in 2026 nuclear- and grid-capacity reporting (OilPrice.com; industry SMR financing coverage).

ExxonMobil executive Dan Ammann put the industry's pitch to Big Tech plainly: gas power, he said, “is low carbon, it's available on an accelerated timeline, and it avoids all the grid interconnection challenges.” Chief executive Darren Woods framed it as the only realistic option on the table today.

“There are very few opportunities in the short term to power those data centers and do it in a way that minimizes, if not completely eliminates, the emissions.”Darren Woods, Chairman & CEO, ExxonMobil

The Arithmetic That Won't Disappear

That framing is contested, and the dispute sits at the center of the story. BloombergNEF is now tracking 99 proposed gas plants tied to data-center projects, representing 126 gigawatts of new generating capacity. Run at industry-standard rates, those plants alone would emit roughly 318 million metric tons of carbon dioxide a year — an increase of about 20 percent over total U.S. power-sector emissions. Run at maximum output, the increase could reach a third. Two projects alone, tied to Amazon and Microsoft sites in Texas, could together emit as much as 45 million metric tons annually — on the order of half of Washington state's yearly total.

What the Buildout Could Add to U.S. Power Emissions

Existing U.S. power-sector CO₂ emissions, plus data-center gas plants under two utilization scenarios

Source: BloombergNEF gas-plant tracking, cited in Bloomberg/Fortune reporting, Aug. 2026 (99 proposed plants, 126 GW).

“There is immense, immense pressure on the whole sector to get power, and get it fast,” David Pomerantz of the Energy and Policy Institute has observed. “They're sort of agnostic if it is clean or dirty.”

“The companies who set the bar for corporate climate action are now bringing new fossil infrastructure online at a breakneck pace.”Drew Wilkinson, former Microsoft sustainability staffer

Carbon capture, still unproven at this scale outside pilot projects, is the industry's answer to that arithmetic. Whether it can be built, permitted and operated fast enough to matter — on the same accelerated timeline that makes gas attractive in the first place — is a question that will not be settled by press release. It is likely to be the central regulatory and reputational battle of this transition.

What It Means for the Digital Oilfield

For an industry already mid-transformation — one increasingly run on drone inspection fleets, digital twins and machine-learning models rather than instinct and drill bits alone — this is not a detour. It is a convergence. The same companies that spent the last five years building sensor networks to monitor methane leaks and structural fatigue on offshore platforms are now the companies best positioned to build, monitor and secure the power plants training the AI models that will, in turn, run those very sensor networks. Chevron's investment in ION Clean Energy's carbon-capture technology, GE Vernova's turbine partnerships, and the rapid build-out of gigawatt-scale, off-grid, single-customer power islands all demand the same digital infrastructure — SCADA systems, predictive maintenance, cybersecurity hardening — that this publication has argued for two years is the real product oil and gas companies now sell.

There is a case, made openly by Exxon and Chevron executives, that this is a net good: gas paired with real carbon capture could deliver power faster and cleaner than an overloaded grid limping along on decades-old interconnection rules. There is an equally serious case, made by climate researchers and former Big Tech sustainability staff, that the industry is using an AI gold rush as cover to build fossil infrastructure that will operate for twenty-year contract terms — long after any individual chatbot fades from memory. Both cases are being tested in real time, on real land, with real capital, in West Texas and beyond.

What is no longer in question is the shape of the industry itself. The oil major of 2026 increasingly looks less like a driller and more like a utility with a drilling division — a company whose most valuable asset may not be what sits beneath its leases, but the gigawatts it can put behind a fence, under a twenty-year contract, for a customer that does not want oil at all. It wants electrons, on demand, starting now. And Big Oil, perhaps to its own surprise, turns out to be very good at delivering exactly that.

◆