Three Companies. One Bottleneck. A Trillion-Dollar Problem.
If you wanted to design the perfect supply-chain chokepoint for the AI era, it might look a lot like the global heavy-duty gas turbine market. There are, for practical purposes, exactly three companies on Earth capable of manufacturing the large-scale turbines that power gas-fired plants at grid scale: GE Vernova, Siemens Energy, and Mitsubishi Heavy Industries. Together, these three account for more than 70% of global production capacity. And right now, demand for what they make is so far ahead of what they can produce that, according to the same Bloomberg investigation, more than $400 billion worth of planned gas-fired power plants through the end of the decade are in jeopardy of delay or cancellation — not because anyone’s short on gas, or capital, or land. Because there simply aren’t enough turbines.
Why Now: It's the Data Centers, Obviously
The proximate cause is the artificial intelligence buildout, and the scale involved is genuinely hard to overstate. Modern hyperscale data centers draw power in the range of 50 to 100 megawatts up to several gigawatts each, and Mitsubishi Power’s own market analysis found that its 10-year forecast for turbines above 100 megawatts has practically doubled in just the last year. The U.S. Department of Energy estimates data center power demand could reach 6.7 to 12% of total U.S. electricity usage by 2028, up from roughly 176 terawatt-hours in 2023 to several hundred terawatt-hours within a few years. Gartner, cited in the same Mitsubishi report, forecasts that 40% of AI data centers will be operationally constrained by electricity deficits by 2027. For “behind the meter” power — generation built directly alongside the data center it serves — gas turbines remain, for now, the most practical option available, given their power density, efficiency, and reliability relative to the alternatives.
The Numbers Behind the Backlog Are Almost Absurd
GE Vernova’s first-quarter 2026 results tell the story in a single line: orders surged 71% year-over-year to $18.3 billion, pushing the company’s total backlog to $163 billion — up from $116 billion just one quarter earlier, according to the company’s reported Q1 results. That backlog represents roughly two and a half years of forward revenue, already contracted. Siemens Energy isn’t far behind: the company disclosed a record €136 billion order book, the largest in its history, with roughly 60% of its 2025 gas turbine orders directly tied to data center projects, according to the Financial Times. According to CompressorTECH²’s reporting on the same dynamic, S&P Global Commodity Insights has tracked wait times for new gas-fired turbine equipment stretching as long as five to seven years, depending on the model and project location.
"It's Cartel Pricing": What This Means for the Price of Building a Power Plant
Scarcity this severe doesn’t just create delays — it reshapes pricing power entirely, and the people actually trying to build plants are noticing. One energy company CFO, put it about as bluntly as anyone in the sector has on record: total project capital costs have doubled in just fifteen months, from roughly $1,000 per kilowatt to more than $2,000 per kilowatt. “It’s cartel pricing,” the CFO said. “There’s nowhere else to go.” The scarcity has gotten strange enough that some developers are reportedly scouring Alberta and the Gulf Coast for refurbished 1990s-era turbines simply because new units aren’t available at any price within their build timeline — and others, including major industrial conglomerates, are buying turbines on spec with no specific project attached yet, purely to guarantee they’ll have equipment in hand when they need it.
Is the Turbine Actually the Bottleneck? GE Vernova Says No.
Here’s a genuinely useful corrective to the panic narrative: not everyone agrees the turbine itself is the binding constraint. Natural Gas Intelligence reported in late May 2026 that a top GE Vernova executive pushed back directly on the framing that turbines are what’s actually slowing data center buildouts, even while confirming that lead times now top three years and the company’s backlog has reached 100 GW. A separate, widely circulated industry analysis made a complementary point: high-voltage transformers, switchgear, and grid-tie batteries — not turbines — may be the truly gating constraint, with pre-2020 transformer lead times of 24 to 30 months now stretching to five years. Electrical equipment, that analysis noted, represents under 10% of total data center cost but effectively 100% of the bottleneck. It’s a useful reminder that in a supply chain this stressed, there’s rarely just one chokepoint — there’s a whole queue of them, and turbines are simply the most visible.
The Manufacturers' Dilemma: Expand, But Not Too Fast
Here’s the genuinely tricky part for GE, Siemens, and Mitsubishi: they’ve all lived through this movie before, and it didn’t end well. Industry analysis cited by Tech Fund pointed directly to the early 2000s, when a boom in orders following U.S. power sector deregulation led suppliers to aggressively expand capacity — only to suffer years of painfully low utilisation once those orders normalised and demand cooled. That history is shaping today’s capacity decisions directly: GE Vernova has been notably cautious about adding new manufacturing capacity, while Mitsubishi Heavy’s CEO Eisaku Ito told Bloomberg his company had been targeting a 30% capacity increase — and concluded, in his own words, “that’s not enough.” GE Vernova has nonetheless committed to expanding production from 55 to between 70 and 80 heavy-duty turbines annually starting in 2026, backed by roughly $300 million in new investment across sites including South Carolina and New York. Across the three manufacturers, overall gas turbine production capacity is on track to climb from roughly 70 gigawatts in 2024 to over 90 gigawatts by 2029 — a meaningful expansion, but one still measured against a demand curve that keeps climbing faster than supply can follow.
Is There an Alternative? Sort Of — But It Comes With Trade-offs
Faced with multi-year turbine queues, some developers are finding workarounds, each with its own cost. Natural Gas Intelligence noted that with large-frame turbines sold out through 2028, some data center developers are pivoting to reciprocating engines instead — a technology that rivals turbines in available manufacturing scale and proved its speed when xAI used reciprocating engines to power up a data center in a matter of months, far faster than a turbine order could have delivered. A more dramatic example came from Oracle: the company announced it was scrapping gas turbines and backup diesel generators entirely for its multi-gigawatt Project Jupiter AI facility in New Mexico, replacing them with a fully islanded microgrid built entirely on Bloom Energy fuel cells — selected, according to the companies involved, not just for speed but also for cleaner local air quality and stronger community acceptance. It’s a meaningful signal: when the default hardware is sold out for years, even the largest buyers in the world start seriously evaluating alternatives that didn’t look commercially competitive just two or three years ago.
What This Means If You're Trying to Build Anything That Needs Power
Constancy Researchers’ honest assessment: the gas turbine bottleneck is one of the cleanest real-world examples available right now of how a single piece of specialised hardware can become the rate-limiting step for an entire technology revolution. Three manufacturers, a backlog measured in hundreds of billions of dollars, wait times stretching toward the better part of a decade, and a structural memory of the early-2000s overcapacity bust that’s actively making those same manufacturers cautious about solving the problem too aggressively. For anyone planning power-hungry infrastructure — whether that’s a hyperscaler, a utility, or an industrial developer — the practical lesson is blunt: if you didn’t lock in your turbine order years ago, you’re now choosing between waiting, paying a premium for someone else’s slot, or seriously evaluating an alternative power source you might not have considered viable as recently as 2023.
Looking to navigate geopolitical risks and identify new industry opportunities?
Connect with Constancy Researchers for data-driven insights, geopolitical intelligence, and strategic consulting tailored to your business needs.
