AI’s Power Crunch: Used Jet Engines Fill the Gap Left by a Six-Year Turbine Backlog
新云社 reports: The artificial intelligence boom has a power problem, and it is getting creative. With new industrial gas turbines effectively sold out for years and lead times stretching to six years, AI data center developers are turning to refurbished power equipment and jet-engine-based generators to keep their expansion plans alive. The scramble underscores how the physical infrastructure underpinning the AI revolution is becoming as scarce as the GPUs that power it.
Why the Turbine Shortage Matters
Hyperscalers and neocloud operators are racing to secure hundreds of megawatts — and increasingly gigawatts — of reliable baseload power for training and inference clusters. Gas turbines have long been the go-to solution for fast, on-site generation, but the global supply chain is straining under simultaneous demand from utilities, grid operators and the AI buildout itself. The result: multi-year waits for new units, soaring prices for used equipment, and a willingness among developers to accept unconventional solutions.
- Lead times: New turbine orders can take up to six years to deliver, far outpacing the 18-24 month construction cycles typical of data centers.
- Refurbished market: Used and remanufactured turbines are commanding premium prices as buyers compete for any available capacity.
- Jet engines: Aeroderivative generators derived from aircraft engines are being deployed as stopgap power sources despite higher fuel costs and emissions concerns.
Implications for the AI and Crypto Compute Economy
This is not just an energy story — it is a compute story. Power availability increasingly determines who can train frontier models and who can operate high-density mining and GPU infrastructure. For decentralized compute networks and on-chain AI agent platforms that promise to aggregate distributed GPU capacity, the bottleneck is shifting from silicon to electricity. Operators that can lock in power — even via unconventional sources — gain a durable competitive edge, while those reliant on grid interconnection face permitting delays and rising costs.
Pollution and permitting are the wild cards. Jet-engine generators are louder and dirtier than modern turbines, and local opposition or air-quality rules could stall projects that seemed viable on paper. Developers are effectively trading regulatory risk for speed.
Forward Outlook
Expect a wave of capital into distributed generation, behind-the-meter power, and long-duration energy storage as AI firms hedge against turbine scarcity. Nuclear small modular reactors and geothermal are drawing renewed interest, but their timelines remain long. In the near term, the used-equipment and aeroderivative market will keep the lights on — and the models training. The broader lesson for the crypto-AI convergence: infrastructure scarcity, not model quality, may define the next phase of competition.








