Jet engine shortages threaten AI data center expansion, wait times into 2030 (www.tomshardware.com)

AI data‑center expansion is bumping up against an unexpected bottleneck: aeroderivative jet turbines. Hyperscalers racing to deploy capacity while waiting years for grid interconnects are buying ex‑airliner cores (GE’s LM6000/LM2500, based on the CF6 family) mounted on trailers for fast‑start bridging power. Orders have surged—Siemens says >60% of its U.S. turbine bookings are tied to AI—and manufacturers are quoting delivery windows into 2028–2030. Major buys include Crusoe/Crusoe Energy’s 29 LM2500XPRESS units (roughly 1 GW) and ProEnergy’s delivery of over 1 GW of PE6000 retrofit systems to two clients. Some customers are paying large deposits or reservation fees to secure slots. That demand is significant because these turbines are modular and fast (LM6000 can reach ~50 MW in <10 minutes; trailer systems can be live in ~30 days), but they’re heavy, long‑lead industrial products requiring high‑alloy castings, precision parts and weeks of testing—constraints that can’t be rapidly scaled. The widespread use of diesel/methane units raises emissions and regulatory pushback (temporary permits often don’t anticipate multi‑year operation), and on‑site generation complicates grid planning and cost allocation. With manufacturers ramping capacity but not until 2028+, the near‑term AI buildout may be limited by fuel‑burning turbines rather than chips, potentially locking in carbon‑intensive solutions while alternative long‑term options (SMRs, hydrogen) remain distant.