AI data centers are using rebuilt aircraft engines as gas turbines to get power sooner, despite higher emissions and lower efficiency.
Some AI data centers are turning to retired aircraft engines as grid connections and power equipment face years-long delays. Companies rebuild the engines into turbines that burn natural gas to generate electricity. TechCrunch reported on the trend November 3, citing The New York Times. The approach can get facilities powered sooner, but it has efficiency and pollution trade-offs.
Why data centers need faster power
The rapid expansion of AI facilities has increased demand for electricity and power equipment. Developers facing long waits for grid connections are looking for on-site generation. The International Energy Agency says a 100-megawatt grid connection in Northern Virginia can take about seven years, according to TechCrunch. That wait can delay a project even after a company has built the data center.
Refurbishing existing engines can get power online sooner than waiting for new equipment. “It’s all about getting power as fast as possible.” — Audun Martinsen, Rystad Energy analyst. The New York Times quoted him on the rush to secure electricity. TechCrunch reported Martinsen estimated these units could power a data center within 24 months. That estimate describes a possible timeline, not a guarantee for every site.
How old aircraft engines become power turbines
These are not aircraft engines left in planes or run on aviation fuel. Suppliers remove engines from retired aircraft, modify them for stationary use and configure them to burn natural gas. The equipment then drives a generator to produce electricity. This type of aeroderivative turbine takes its design from an aircraft engine. The units can be installed on site, close to a data center’s electrical load.
TechCrunch reported that ProEnergy rebuilds a GE engine commonly used on Boeing aircraft. The company also sells newly manufactured PE6000 aeroderivative units. In April, Crusoe announced an order for 13 such units for planned AI data-center projects. ProEnergy said each unit can generate 50 megawatts, for a combined 650 megawatts. The order was scheduled for delivery by summer 2027, so it should not be described as already installed capacity.
Companies using on-site gas generation
Elon Musk’s xAI is already using converted aircraft-engine turbines at its Colossus 2 data center in Memphis, TechCrunch reported. The New York Times said xAI had installed eight units in Mississippi, citing state permits and satellite imagery. The turbines are intended to help supply the facility’s electricity. These details come from the cited reporting, not a public operating report from xAI.
Other developers are also buying gas turbines for data centers, though not every project uses recycled aircraft engines. OpenAI’s first Stargate site in Abilene, Texas, has gas turbines. Crusoe has separately ordered ProEnergy’s converted aircraft-derived units. Keeping these examples distinct matters because “gas turbine” includes several kinds of machines with different designs and delivery schedules.
Power speed comes with environmental costs
The attraction is speed, but converted aircraft engines are not the most efficient option. TechCrunch says they are less efficient and more polluting than turbines designed for conventional gas plants. Reuters Events says small open-cycle turbines also emit more greenhouse gases per megawatt-hour than larger combined-cycle plants. Combined-cycle plants recover heat to generate extra power.
That difference can affect costs and emissions over time. Reuters cited the Electric Power Research Institute: smaller turbines’ capital costs per unit of electricity can be up to 50% higher. The comparison concerns smaller turbine categories broadly, not every converted engine. Developers may still accept higher costs when grid power or new equipment is years away.
What the turbine shift signals
The move shows how delays in electricity supply are shaping data-center construction. Reuters Events estimates U.S. behind-the-meter gas generation will grow by 29.6 gigawatts through 2030, with data centers accounting for roughly 88% of demand. That forecast covers gas generation overall, not only converted aircraft engines. It points to wider use of on-site power while grid capacity catches up.
Reuters also reported that manufacturing shortages have stretched delivery times for large gas turbines. Smaller and aircraft-derived units can arrive sooner, which helps explain their appeal to developers. Yet faster installation does not erase their higher fuel use and emissions compared with more efficient plants.
Ultimately, retired aircraft engines offer some data centers a faster path to electricity, not a clean or permanent fix. Developers must weigh the time saved against fuel use, emissions and operating costs. Grid upgrades and newer power plants may change the choices available to future projects. The race to power AI facilities is likely to keep testing those trade-offs.
