AI data centres turn to used jet engines as new turbines take six years

Key points

• Crusoe ordered 13 jet-engine-derived turbines from ProEnergy in April, rated at 50 megawatts each for 650 megawatts in total.
• Global turbine orders have topped 100 gigawatts against annual manufacturing capacity of 60 to 70 gigawatts.

AI data-centre developers are turning to used gas turbines and retired jet engines because new power equipment can take years to arrive, according to Reuters. New gas turbines now average six years for delivery in the US, energy consultancy Wood Mackenzie says, with talks already centred on 2030 and 2031. That timing clashes with an industry trying to bring large computing campuses online far sooner.

Global turbine orders have exceeded 100 gigawatts, against manufacturers' estimated annual capacity of 60 to 70 gigawatts.

One workaround is the aeroderivative turbine, which adapts aircraft jet-engine technology for stationary power. Second-hand machines also sell at a discount: used turbines in good condition fetch about $1.5m to $1.7m per megawatt, against roughly $2m to $2.5m for new ones, brokers and analysts told Reuters.

Second-hand power brings complications. On-site plants need fuel, face air-emission rules and require permits, and the Environmental Protection Agency says data-centre turbines and engines fall under federal air-pollution standards. Cleanview found about 90 gigawatts of proposed on-site generation at US data centres in its mid-2026 analysis, with only about 2 gigawatts operating.

Read the full story on Decrypt →