AI data centers are burning gas in converted jet engines to skip the grid queue

AI data centers are burning natural gas in rebuilt jet engines because the grid cannot connect them fast enough. The New York Times reported on November 3, 2026 that operators waiting on grid hookups and new turbines are buying retired aircraft engines converted to run on gas. It is a fast, dirty workaround, and it shows how far the race for compute has outrun the power system.

What is being installed

TechCrunch's Aditya Mehta, summarizing the Times report, says the engines are retired aircraft units rebuilt to burn natural gas. They are less efficient than conventional gas-plant turbines and pollute more, so more units are needed for the same output.

The main example is xAI. According to state permits and satellite imagery cited by the Times, the company installed eight converted jet-engine turbines in Mississippi to help power its Colossus 2 data center in Memphis. The units come from ProEnergy, which rebuilds a GE engine commonly found on Boeing jets. We reported on the legal fight over xAI's earlier turbines in our piece on the NAACP Clean Air Act suit.

xAI is not alone. TechCrunch reports that OpenAI followed in 2025 with gas turbines at its first Stargate site in Abilene, Texas, built by Crusoe. Crusoe ordered 13 ProEnergy units in April.

The numbers behind the order

ProEnergy's own press release, dated April 16, 2026, gives the Crusoe order in detail: 13 PE6000 aeroderivative gas turbine generator sets, each producing 50 megawatts, for 650 megawatts in total, with delivery in summer 2027. The equipment is to be test-fit and quality-checked at ProEnergy's Sedalia, Missouri facility before shipping.

In the release, Crusoe's senior vice president of power infrastructure, John Adams, said: "At Crusoe, we are committed to leading the industry with an energy-first approach to AI infrastructure." ProEnergy's chief executive, Jeff Canon, said: "Our equipment is made in-house for world-class reliability no matter the application."

Crusoe's other power plans have been less smooth. TechCrunch reported on September 25, 2026 that Crusoe abandoned a $1.25 billion plan to buy 29 of Boom Supersonic's 42-megawatt Superpower turbines, with first deliveries slated for 2027. A Crusoe spokesperson, Andrew Schmitt, said the company's energy plans "haven't changed" and that it still plans to use turbines, "just not Boom's." The same report says the Abilene campus built for Oracle and OpenAI is grid-powered with a gas plant for backup only, while a 900 megawatt Abilene site for Microsoft uses on-site gas turbines.

Boom's chief executive, Blake Scholl, wrote that the company "will be delivering about 250MW of Superpowers next year to other sites, and we're targeting 1GW in 2028."

Why the grid is the bottleneck

TechCrunch cites the International Energy Agency for the core fact: a 100 megawatt grid hookup in Northern Virginia takes about seven years. We could not open the IEA report this run, so that figure is attributed to TechCrunch's summary of it.

Against that, Rystad Energy analyst Audun Martinsen told the Times that jet-powered units let a data center run in about 24 months. His words: "It's all about getting power as fast as possible," and "they can have a data center running in 24 months."

Turbines are also scarce. TechCrunch cites a Bloomberg report that Mitsubishi Power expects gas turbine demand to push deliveries out to 2031, with most capacity coming by then. That pushes buyers toward used or converted engines, which are available sooner.

The Microsoft case shows the other side of the problem. Microsoft pulled out as anchor customer for Nscale's data center in Loughton, England, after the local utility said it could not supply enough power until the early to mid 2030s. We covered the numbers in our report on the Loughton agreement ending.

The policy backdrop

TechCrunch notes that the Trump administration is encouraging data centers to generate their own electricity. That policy turns the workaround into a preferred path in some places. But TechCrunch also points to the cost: the converted engines are less efficient and pollute more than conventional gas-plant turbines. Gas generation near homes has already drawn a Clean Air Act lawsuit from the NAACP against xAI, which survived a motion to dismiss on October 30, per our earlier report. Local opposition to data centers is rising, as our data center backlash tracker shows.

What the reporting does not tell us

Several gaps are worth stating plainly.

• No megawatt figure for xAI. The Times report, as summarized by TechCrunch, gives no output for the eight Mississippi units.
• No emissions numbers. Both outlets say the engines pollute more, but neither article we opened gives a measured figure.
• Secondhand sourcing. The Times article is paywalled. We rely on TechCrunch's summary of it, plus ProEnergy's and Boom's own statements, which are vendor material.

Our take

The jet engine story is a clean marker of where the race has hit a physical limit. When labs will accept more pollution and a more expensive power source to save years, the constraint is no longer chips or money. It is electricity and the permits to deliver it. That changes who wins: a lab with a signed power supply can train sooner than a richer rival stuck in a queue, whatever its chips cost.

What we would watch next: whether permits for converted units face challenges the way xAI's turbines did, whether turbine makers add capacity before 2031, and whether any hyperscaler commits to a data center with no on-site gas at all. The 24 months versus seven years comparison explains the choice. It does not make the choice cheap.