AI data centers are burning jet engines for power

The wait for grid power has gotten long enough that AI data centers are generating their own, and some are doing it with retired jet engines. As TechCrunch reports, drawing on the New York Times, companies are buying aircraft engines rebuilt to burn natural gas because they can be installed far faster than a new grid connection or a purpose-built gas turbine. xAI put eight converted turbines in Mississippi to help power its Colossus 2 site in Memphis, using units from ProEnergy that are rebuilt from a GE engine common on Boeing jets. OpenAI did something similar at its first Stargate site in Abilene, Texas.

The numbers explain the appeal. Connecting 100 megawatts to the grid in Northern Virginia takes about seven years, one analyst told the paper, while a converted engine can have a data center running in 24 months. A Mitsubishi Power executive said most new conventional power capacity is not expected until 2031. The squeeze is real enough that Microsoft walked away as the anchor customer for a data center in Loughton, England, after the local utility said it could not supply enough power until the early-to-mid 2030s.

The catch is that these engines were built for airplanes, not power plants. They are less efficient and pollute more than purpose-built turbines, so operators need more of them to produce the same power. Several experts in the article expect this to be a short-term fix as efficiency and emissions concerns grow.

Why it matters

If you follow the economics of AI, power, not chips, is becoming the binding constraint, and the workaround carries visible costs. For anyone living near these sites, it means more local gas generation and emissions, decided outside the usual utility planning.