Share
The AI boom has a hardware bottleneck that has nothing to do with chips. It is gas turbines, and there are not nearly enough of them.
Utilities, developers and companies building power for AI data centers are hunting worldwide for second-hand turbines, according to Reuters. They are bringing mothballed machines back to life and turning retired jet engines into power generators. The reason is simple: new turbines in the US now take six years on average to deliver, an analyst told the outlet.
The waitlist problem
The supply crunch shows up clearly in the numbers. Global turbine orders have climbed past 100 GW. Annual manufacturing capacity sits at just 60 to 70 GW.
Many manufacturing slots for future orders are already spoken for through 2030 and 2031. That pushes anyone arriving late into the next decade, which is awkward when your data center opens next year.
The big three turbine makers are feeling the pull. Siemens Energy, GE Vernova and Mitsubishi Power have all reported unprecedented backlogs, with AI-related electricity needs making a significant dent in their order books.
The demand driver is not mysterious. US data centers consumed nearly 17 GW of power in 2023, and that figure could potentially more than triple by 2028. Projections suggest US gas generation capacity tied to data centers may reach approximately 29.6 GW by 2030, with 88% of that growth driven by data centers.
A booming market for hand-me-downs
Turbines are now playing out a similar script. Estimates show well-maintained second-hand units fetching between $1.5 to $1.7 million per MW. Refurbished units cost around $1.3 million per MW.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
We respect your privacy. Unsubscribe anytime.
New turbines, for comparison, run $2-2.5 million per MW. So used equipment still comes in cheaper on paper, but the gap is narrowing as buyers bid against each other. The real premium is not on price. It is on time.
Some developers are skipping the large-frame turbine queue entirely. Crusoe, the data center developer, ordered 13 aeroderivative units from ProEnergy for approximately 650 MW of AI projects in the US.
Aeroderivative generators are essentially jet engine technology adapted for the ground. They are smaller and quicker to deploy than the giant industrial turbines that anchor traditional power plants.
What this means
For turbine makers and service firms, the picture is broadly favorable. Research findings suggest both original equipment manufacturers and companies that refurbish and maintain turbines stand to benefit from higher revenues. Long backlogs give manufacturers pricing power and visibility.
For data center operators, the outlook is less comfortable. Rising competition for turbines may push long-term power costs higher for data centers, which could squeeze operating budgets and margins.
There is also a strategic risk in leaning harder on natural gas. Heavier reliance on gas, combined with constrained manufacturing capacity, could nudge companies toward alternative power sources. That pressure may be strongest in regions that depend on fuel imports or face emissions-related policy limits.
Second-hand equipment carries its own trade-offs. A refurbished turbine can arrive fast, but older machines may come with higher maintenance needs and less efficiency than new models.
What to watch next: whether Siemens Energy, GE Vernova and Mitsubishi Power expand production quickly enough to shrink that six-year wait, and whether used turbine prices keep climbing toward the cost of new units.
