Grounded jet engines take off again as datacenter generators (www.theregister.com)

Soaring AI compute demand has created a squeeze on stationary gas turbines used for on-site power, prompting some operators to repurpose retired aircraft engines as temporary datacenter generators. Major turbine makers (MHI, Siemens, GE Vernova) supply roughly two‑thirds of the market and are reporting multi‑year backlogs—three to five years for larger units—and even non‑refundable reservation fees (one developer reportedly paid $25M for a 2030 slot). With hyperscalers expecting US datacenters to draw ~22% more grid power by end of 2025 and grids in many regions unable to keep pace, firms like ProEnergy are buying, overhauling and converting GE CF6 jet engines into PE6000 turbines to drive generators during construction and early operation; these are typically demoted to backup once grid power arrives. For the AI/ML community this is consequential: infrastructure rollouts can be delayed, capex and procurement complexity rises, and operators may rely more on higher‑emission diesel or bespoke fixes that affect sustainability targets. Technical implications include engineering work to adapt high‑RPM turbofans for continuous shaft‑power duty, different maintenance regimes, permitting and acoustic considerations, and uncertain long‑term economics compared with fuel cells, SMRs or grid upgrades. The jet‑engine workaround is a practical stopgap, but it underscores broader supply‑chain and energy‑capacity constraints that could shape where and how AI capacity is deployed over the coming years.