Data centers hunt for used gas turbines and converted jet engines as new units slip to 2030

AI & Compute

Gas Turbine Orders Top 100 GW as AI Power Supply Tightens

Worldwide gas turbine orders exceed 100 GW against 60-70 GW of annual production, as data centers turn to used frames, aeroderivatives and reciprocating engines to bridge a six-year delivery gap.

By
Rebecca Stone
Filed
Channel
AI & Compute
Read
4 min read

By Rebecca Stone

Worldwide orders for gas turbines exceed 100 GW, against estimated production capacity of 60-70 GW a year, Reuters reported, as utilities, developers and data center companies hunt for equipment to avoid multiyear delivery delays.

In the United States, a new gas turbine takes about six years to deliver on average, and a Wood Mackenzie analyst said delivery discussions now center on 2030-2031.

Jefferies estimates U.S. data-center turbine demand at nearly 17 GW this year. It could more than triple by 2028, when the firm sees 87 GW of global manufacturing capacity against 129 GW of demand.

Can buyers sidestep the factory queue?

Utilities, developers and data center companies are buying second-hand turbines, restarting mothballed machines and converting retired jet engines because newly ordered equipment will not arrive until the next decade.

Reuters reported that well-maintained used turbines sell for about $1.5 million to $1.7 million per MW, refurbished units for about $1.3 million per MW, and new equipment for $2 million to $2.5 million per MW.

ReflowX general manager Dan Poole said two 14.5 MW Solar Turbines units sold for $12.5 million combined; equivalent new machines would have cost about $35 million and taken five years to deliver.

“That secondary market is very, very hot right now,” EthosEnergy CEO Ana Amicarella told Reuters, adding that prices being demanded have escalated. Darayus Pardivala, who runs Sulzer’s U.S. aftermarket business, said his team receives a request to find a used turbine for a data center every few weeks.

The Tennessee Valley Authority said prices for new, used and refurbished equipment have increased significantly over the past two years, while the National Rural Electric Cooperative Association said some member utilities are turning to refurbished turbines.

Can jet engines close the gap faster?

Aeroderivative turbines use aircraft engine cores. They are smaller and less efficient than the heavy-duty frame turbines utilities use for baseload plants, but they can be built and moved faster.

In April, Sedalia, Missouri-based ProEnergy agreed to supply Crusoe with 13 PE6000 aeroderivative generator sets rated at 50 MW each, 650 MW in total, for Crusoe’s U.S. AI data center projects. Reuters said the units are based on GE Aerospace jet engines.

xAI’s Mississippi power site is permitted for 41 gas turbines, including eight 50 MW aeroderivative units. SVI Bremco sales director Scott Schreeg said some developers deploy more than 100 smaller turbines at a single site, mixing models and brands.

Reciprocating engines are filling another part of the gap. Wärtsilä said it has booked about 3 GW of gas engine power unit orders from U.S. data centers since spring 2025, and the Financial Times reported that Anthropic last week paid $3.4 billion for 1.7 GW of Caterpillar generators that another buyer had ordered.

Some developers are skipping combustion. Google’s planned Cheyenne, Wyoming campus will run on 702 MW of fuel cells, while MLQ tracks on-site and behind-the-meter power for AI campuses in its AI infrastructure research.

Can turbine factories expand fast enough?

The three dominant large-gas-turbine makers—GE Vernova, Siemens Energy and Mitsubishi Heavy Industries—all say AI-driven demand is outrunning what they can build.

GE Vernova reported that its gas power equipment backlog and slot-reservation agreements grew to 122 GW in the third quarter from 116 GW; the company expects at least 130 GW by year-end 2026. During the quarter, it signed 11 GW of new gas contracts, including 8 GW of slot reservations, and shipped 5 GW of equipment.

Noor Qureshi, an associate at Enverus Intelligence Research, said limited excess supply makes it “unlikely to offset a meaningful share of the delays.” Used and converted machines can carry unclear maintenance records, limited remaining operating life and thin long-term service support.

Converted aircraft engines also raise questions about their durability under round-the-clock operation. Even after buyers find hardware, air permits and grid connections can still delay a project.

Until factory capacity rises above the estimated 60-70 GW a year and U.S. delivery windows move earlier than 2030-2031, data-center buyers are likely to keep competing for scarce machines, keeping used and refurbished prices elevated.

Source: mlq.ai

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

744 articles

Related articles

« Previous articleNext article »