Retired aircraft engines are being rebuilt to generate electricity for data centers. The approach could bring power online faster, although it carries efficiency, pollution and fuel-supply costs.
AI data centers are turning to converted aircraft engines and natural gas turbines while grid connections remain years away.
Rystad Energy analyst Audun Martinsen told the New York Times, as reported by TechCrunch on November 3, that converted jet-engine generation could allow a data center to operate within about 24 months.
TechCrunch contrasted that estimate with an approximately seven-year wait for a 100-megawatt grid connection in Northern Virginia, citing the International Energy Agency. The seven-year figure is a regional benchmark, not a measured comparison involving the same project.
The attraction is straightforward: a developer that can secure its own electricity may begin operating before the utility grid is ready. Whether that advantage survives fuel, equipment and permitting constraints remains uncertain.
A faster route to power
The systems use engine cores recovered from retired aircraft, rebuilt to burn natural gas and drive electricity generators.
TechCrunch, drawing on New York Times reporting, said xAI installed eight converted jet-engine turbines in Mississippi to help power its Colossus 2 data center in Memphis. The reporting cited state permits and satellite imagery, although those records do not independently establish when all eight units began delivering electricity.
ProEnergy's PE6000 uses a CF6-80C2 engine core found in aircraft including the Boeing 747. The company's 2023 product announcement rated the turbine at 48 megawatts and said it could serve about 40,000 homes.
The appeal is schedule control. Developers can order generating equipment and build on-site capacity while waiting for a utility interconnection, transmission upgrade or new substation.
That can shift the timeline for delivering computing capacity away from the utility's construction schedule. It does not eliminate the need for natural gas supply, environmental approval or reliable electricity infrastructure.
The projects already in motion
OpenAI's first Stargate site in Abilene, Texas, developed by Crusoe in 2025, uses gas turbines, as described in TechCrunch's November report.
An important distinction is that the original Oracle and OpenAI campus is grid-powered, with gas turbines serving as backup generation. Crusoe is separately developing another Abilene data-center project for Microsoft that is intended to use on-site gas turbines, according to TechCrunch's September reporting.
In April 2026, Crusoe ordered 13 ProEnergy turbine generators for its AI factories.
ProEnergy's April 16 announcement rates each unit at 50 megawatts, giving the order 650 megawatts of combined generating capacity. Its earlier PE6000 announcement listed 48 megawatts per unit. The available releases do not explain the difference.
The equipment order confirms a commitment to on-site generation, not that the turbines have been commissioned or are already supplying 650 megawatts.
Grid delays have also affected commercial agreements. Microsoft withdrew as anchor tenant for Nscale's Loughton data-center project in England after electricity availability was pushed back, TechCrunch reported.
That decision illustrates the commercial cost of waiting. A delayed project can lose a major customer before the grid arrives.
Speed brings an efficiency penalty
Converted aircraft engines offer a potential shortcut to power, but they are less efficient and more polluting than some conventional generation options.
Simple-cycle turbines generally use more fuel per unit of electricity than more efficient combined-cycle gas plants, which recover exhaust heat to generate additional power.
For data-center operators, the choice can mean accepting higher operating costs and emissions in exchange for earlier access to computing capacity.
Natural gas generation also exposes developers to fuel prices, pipeline access, supply logistics and local air-permitting requirements.
The International Energy Agency's 2026 analysis estimates that reliable on-site gas generation for data centers may require generating capacity 30% to 70% above demand. The agency cautions that equipment shortages and other constraints mean on-site generation is not necessarily a faster solution at scale.
Commercial commitments can change, too. Crusoe abandoned a separate proposed $1.25 billion agreement for 29 Boom Supersonic turbines in September, while maintaining that it still intended to use turbines from other suppliers, TechCrunch reported.
These risks make converted engines a potential bridge around grid delays rather than a complete solution to long-term electricity demand.
Developers still need dependable generation, permits and, where grid service remains part of the plan, transmission capacity.
The strategy shifts the bottleneck. The question moves from "When will the grid connect?" to "Can the site secure fuel, emissions approval and reliable on-site generation?"
What settles it
The next evidence will come from operating permits, turbine deliveries, commissioning records and the dates when data centers begin receiving usable electricity.
Equipment orders and installed nameplate capacity alone will not establish that converted engines have accelerated commercial operations.
If developers bring computing capacity online before their grid connections are ready, converted engines will have demonstrated their scheduling value.
If gas infrastructure, emissions restrictions or equipment availability delay commissioning, one infrastructure bottleneck may simply have replaced another.
The critical test is not how many jet-derived turbines developers order. It is whether those turbines deliver reliable power sooner, and at an acceptable economic and environmental cost.