Some AI data centers are generating their own electricity with jet engines pulled from retired aircraft. TechCrunch reported on Nov 3, citing a New York Times story, that operators stuck in years-long lines for grid connections and new turbines are rebuilding those engines to burn natural gas and installing them on site.
The practice says a lot about where AI infrastructure stands today. Chips and models get most of the attention, but power is now one of the main limits on how fast new capacity can come online. This explainer covers what the converted engines are, who is using them, what the trade-offs look like, and why the story matters to founders building on top of this stack.
What converted jet engines are and how they work
Aeroderivative turbines are built from the cores of aircraft engines. According to Tom's Hardware's 2025 report, which cited IEEE Spectrum, some of the units use General Electric's CF6-80C2 and LM6000 cores, the same designs found on aircraft such as the 767 and the Airbus A310. Vendors including ProEnergy and Mitsubishi Power package them into modular generators that can deliver about 48 megawatts apiece.
The appeal is speed and modularity. These generators arrive on trailers, start quickly and can be added in blocks as a site grows. They have been used for decades in military and offshore drilling work, and Tom's Hardware described their appearance at data center sites as a sign of how tight power supply has become.
Why data centers are waiting on power
The core problem is time. The International Energy Agency estimates that connecting 100 megawatts to the grid in Northern Virginia takes about seven years, according to TechCrunch. An executive at Mitsubishi Power told Bloomberg in late October that most new power capacity is expected to come online by 2031.
Against that timeline, converted engines look fast. Rystad Energy analyst Audun Martinsen told the Times that these units can have a data center running in 24 months. The delays are already affecting business in some cases. In October, Microsoft pulled out as the anchor customer for Nscale's data center in Loughton, England, after the local utility said it could not deliver enough power until the early to mid-2030s.
Who is using them
The Times reported that xAI has installed eight converted jet-engine turbines in Mississippi to help power its Colossus 2 data center in Memphis, with the units coming from ProEnergy, which rebuilds a GE engine commonly found on Boeing jets. In 2025, OpenAI used gas turbines at its first Stargate site in Abilene, Texas, which Crusoe is building. Crusoe, based in Denver, ordered 13 of ProEnergy's converted units in April.
Crusoe has also been in the news for its funding. Crunchbase's Q3 report listed the AI infrastructure company's $3.9 billion round among the largest late-stage deals of the quarter. In September, according to TechCrunch, Crusoe walked away from a $1.25 billion deal to buy 29 natural gas turbines from Boom Supersonic, a startup developing a supersonic jet that has also begun selling a version of its engine as a data center power plant. A Crusoe spokesperson said at the time that the company still plans to use turbines.
The trade-offs of jet-derived power
Speed comes at a cost. TechCrunch notes that converted jet engines are less efficient and pollute more than turbines designed for conventional gas plants, and that it takes more of them to make the same amount of power. Tom's Hardware explains that aeroderivative units typically run in simple-cycle mode, which burns fuel without capturing waste heat, and that many require emissions controls to meet nitrogen oxide limits.
That may make them a short-term play. Some experts quoted by TechCrunch expect efficiency and environmental concerns to matter more in a year or two, which would push operators toward other sources as grid connections and new equipment arrive. For now, they are a bridge, and the people building data centers are treating them that way.
What it means for AI founders in the Bay Area
If you build AI products, the power queue shapes your costs and your timelines even when you never touch a turbine. Compute supply influences what you can train, how fast you can ship and what you pay, and it feeds into stories like our look at AWS capacity being sold out for AI startups. For a wider view of what else is moving, see our guide to the biggest AI trends of 2026.
Infrastructure is also a topic worth talking through in person. South Bay Renaissance, which Tribe organizes to rebuild the startup and social scene across the South Bay and Peninsula, held an AI Infra Dinner in Palo Alto on Sep 30. The event page described a small group of founders, engineers and researchers working across inference, chips, agent infrastructure and foundation models. If you want to meet the people working on this layer, our list of the best AI communities in 2026 is a good starting point.
Founders can also ask practical questions of their own cloud and hardware providers: how much capacity is reserved, where it is located and what the power plan looks like. Those answers say more about delivery risk than a spec sheet does, and they are worth asking before you sign a long commitment.
Join the AI conversation
Follow the Tribe Events calendar on Luma for upcoming AI dinners and talks in the Bay Area, and join the AI Trends Tribe to talk through stories like this one with other people following the field.
