NewsWhy AI Data Centers Are Buying Used Gas Turbines
A scramble for electricity is pushing U.S. AI infrastructure developers toward refurbished turbines and retired aircraft engines. The equipment can shorten some procurement delays, but it does not remove permitting, maintenance or grid constraints.
Source-linked desk reporting, not firsthand equipment inspection. AI-assisted drafting may be used; see our editorial and corrections policy.
Representative licensed photograph of a power plant; it does not show a turbine purchased by an AI developer. Photographer: Kindel Media / Pexels.
Why are U.S. data centers looking for used turbines?
AI computing campuses require large, dependable electricity supplies. A data center can be built faster than the grid connection or generating plant needed to run it. Reuters reports that manufacturers face global gas-turbine orders exceeding 100 gigawatts against estimated annual production capacity of 60–70 gigawatts. These figures describe industry-wide demand and capacity, not a forecast that every U.S. project will wait the same length of time.
The Reuters report says average U.S. delivery times for new turbines are about six years, with some negotiations stretching into 2030–31. That makes the secondary market attractive to developers seeking equipment earlier, even when the machine requires refurbishment and a new site-specific engineering assessment.
What can developers buy—and what are the trade-offs?
| Route | Potential advantage | Important limitation |
|---|---|---|
| New gas turbine | New equipment with defined specifications | Long manufacturing queue and project lead time |
| Used or refurbished turbine | May be obtainable before a new-build slot | Unknown wear, overhaul costs, remaining operating life |
| Converted aircraft engine | Potentially reusable generating hardware | Conversion, fuel, emissions and continuous-duty suitability |
| Modular gas engines | Phased deployment across multiple units | Fuel supply, permits and delivery still required |
This is an editorial comparison of engineering and procurement considerations, not a ranking of actual suppliers or a claim that used machines are universally cheaper. Developers must assess each project’s load, air-quality obligations, fuel access and reliability requirements.
What do power-equipment companies confirm?
Wärtsilä said on October 5 that it had booked a 282-megawatt U.S. data-center power project using 15 gas engines, with equipment delivery scheduled for 2028 and full operation expected in 2029. The company said its U.S. data-center orders had passed 3 GW. These are manufacturer-reported bookings and plans, not proof that the equipment is already producing electricity.
In its second-quarter 2026 results, GE Vernova reported 116 GW of gas-power equipment backlog and slot-reservation agreements, and outlined plans to expand annual turbine output. Those commitments help explain why ordering a new machine does not necessarily deliver immediate power. The GE Vernova figure combines firm equipment backlog with slot reservations and should not be treated as completed installations.
Will refurbished turbines solve the AI electricity shortage?
Not on their own. The used-equipment market is finite, and equipment may need inspections, replacement parts, controls work and environmental approvals. A generator also needs gas supply, cooling, site works and safe electrical integration. Reuters describes developers exploring retired jet engines as well as mothballed industrial machines, but says secondary-market supply can only partially ease the shortage.
Local residents and regulators may also weigh emissions, noise, water use and whether privately operated generation helps or complicates wider grid planning. An equipment purchase is therefore not equivalent to a permitted, operating data-center power plant.
What should U.S. communities and business customers watch next?
For a proposed campus, the useful questions are concrete: Has the developer secured a generating unit or only reserved a manufacturing slot? Is there an air permit? Who supplies the fuel? Is the facility designed to connect to the grid, operate independently, or both? Has an actual in-service date been disclosed? Public utility filings and local permit records can provide more project-specific answers than national shortage estimates.
For related context, see NewsTrendus coverage of local U.S. data-center restrictions and AI infrastructure's energy costs. Those articles address different policy and environmental questions rather than equipment procurement.
Sources and reporting limitations
- Reuters, November 9, 2026 — reporting on used turbines, retired jet engines, production capacity and delivery queues.
- Wärtsilä, October 5, 2026 — primary announcement of a 282-MW U.S. project and projected timeline.
- GE Vernova, July 2026 — company-reported backlog, slot reservations and manufacturing plans.
- Data Center Dynamics, November 2, 2026 — independent sector reporting on more recent equipment demand.
Original publication: November 9, 2026. This report synthesizes attributed public reporting and company disclosures; no turbine purchase, shipment or permitting record was independently inspected by NewsTrendus.