Cutting corners: The jet engine – once a symbol of speed and flight – is finding new life on the ground, fueling the world's artificial intelligence infrastructure. Faced with long delays in connecting to electric grids and limited access to large gas turbines, data center developers are increasingly deploying aeroderivative turbines and diesel generators to keep AI systems running.
Manufacturers of turbines originally designed for aircraft are now reporting a surge in orders as hyperscale developers seek stopgap ways to secure electricity. "The incentives have never been greater for any sort of technology that can supply power," Kasparas Spokas, director of the Clean Air Task Force's electricity program, told The Financial Times.
Grid connection queues stretching as long as seven years have left developers with few alternatives. While grid power remains cheaper and cleaner, the backlog – combined with growing public discontent over data centers' impact on local utility bills – has made on-site generation a pragmatic choice. By installing turbines and generators directly beside their server farms, companies can continue training and operating AI models without waiting for new transmission capacity.
GE Vernova is among the suppliers seeing an uptick in demand. The company is providing data center developer Crusoe with aeroderivative turbines expected to supply nearly one gigawatt of power to the Stargate facility in Texas, a joint project by OpenAI, Oracle, and SoftBank.
GE Vernova Chief Financial Officer Ken Parks told investors in December that the company was experiencing "growing demand" for its aeroderivative and smaller gas units, which "serve as bridge power supporting data center needs." Orders for these turbines rose by roughly one-third in the first three quarters of 2025 compared with the same period a year earlier.
ProEnergy, another supplier, has sold more than 1 gigawatt of its 50-megawatt gas turbines adapted directly from jet engines. The machines incorporate CF6-80C2 cores – engines that once powered Boeing 747s – though the company is now building more components from scratch.
Boom Supersonic, an aviation start-up backed by Sam Altman, has agreed to sell turbines to Crusoe capable of delivering about 1.2 gigawatts of electricity. The units are "virtually identical" to those built for Boom's supersonic jets. "Three or four years ago, I imagined we would do the airplane first and energy second," Chief Executive Blake Scholl said. "But then I got a call from Sam Altman who said, 'Please, please, please make us something.'" Boom expects turbine sales to help finance its aviation program.
Diesel generation is also making a comeback. Cummins, a major manufacturer of industrial generators, has sold more than 39 gigawatts of capacity to data centers and nearly doubled its output this year. While such systems have traditionally provided backup power, the company's data center executive director, Paulette Carter, said Cummins is seeing "growing interest in on-site primary power."
US Energy Secretary Chris Wright told Fox News in November that officials were exploring ways to use these existing units to shore up the grid. "We will take backup generators already at data centers or behind the back of a Walmart and bring those on when we need extra electricity production," he said.
But reliance on fossil-based energy carries substantial environmental costs. Small on-site units are typically less efficient than grid-scale gas turbines or renewable energy sources. Regulators in several states are beginning to relax restrictions on generator use to accommodate rising data center demand.
In Northern Virginia – the region often called Data Center Alley – the Department of Environmental Quality is considering allowing diesel generators to operate for longer periods. The US Environmental Protection Agency has also signaled that such generators could help stabilize local power supplies.
Even as developers race to secure electricity, the financial trade-offs are significant. On-site generation lacks the economies of scale that make utility power cheaper. Analysts at BNP Paribas modeled the cost of electricity from a small behind-the-meter gas plant that Williams is building in Ohio, where Meta will be a customer. They estimated a price of $175 per megawatt-hour – roughly twice the average cost paid by industrial users.
Image credit: The Financial Times
AI data centers are turning to jet engines and diesel because the grid can't keep up
