The Gas Turbine Bottleneck and the Race for the Next Electron

Electricity costs are rising fast. One midstream energy CFO in Texas told us his company’s rates have climbed from $0.35/MWh in 2021 to nearly $0.70MWh today. “This is getting ridiculous,” he said. “We used to say the price at the pump decides elections. Going forward, it’s going to be the price at the meter.”

Similarly, an executive responsible for the data center buildout at Microsoft asks, “If every producible watt is already committed for the next five years, where can I get more watts?”

That question has become the defining obsession of this moment. AI companies, hyperscalers, and industrial manufacturers are all competing for the same scarce resource: firm, dispatchable electricity. Yet, turbines, the single most important piece of equipment required to make that power, are suddenly the most scarce.

There are effectively only three relevant manufacturers that can produce large-scale gas turbines: GE Vernova ($162B market cap), Siemens Energy ($91B), and Mitsubishi Heavy Industries ($84B). These machines sit at the center of the modern power plant. They are the hardware that converts fuel into motion and motion into electrons. Today, they are booked solid for years. GE, Siemens, and Mitsubishi are each reporting order backlogs that stretch roughly five years, with many delivery slots already sold into the next decade.

Rumors are circulating in Washington about just how acute this has become. A contact close to the Department of Energy’s new venture fund (a kind of “In-Q-Tel for energy”) told us the administration is even considering whether to pressure allies with outstanding GE turbine orders to release them back to U.S. buyers. It sounds extreme, but that’s the level of urgency building around this bottleneck.

What a Gas Turbine Actually Is

A gas turbine is, at its core, an air compressor, a combustion chamber, and a spinning shaft. Air is compressed, mixed with fuel (usually natural gas) and ignited. The expanding gases turn the turbine blades, which drive a generator to produce electricity. In a simple-cycle setup, that’s the whole process. In a combined-cycle plant, the hot exhaust is captured to make steam and power a second turbine, pushing overall efficiency and reducing emissions.

These are enormous machines. A single heavy-duty turbine can weigh more than 400 tons, stretch nearly 50 feet long, and generate anywhere from 100 to 400 megawatts, enough to power a mid-sized city. Smaller aeroderivative models, adapted from jet engines, are used for distributed or peaking power. Each large turbine costs roughly $50-100M depending on its class and configuration. It is elegant, brutally engineered machinery, and the workhorse of global baseload power.

How We Got Here

The roots of the shortage go back two decades. In the early 2000s, gas plants were being built everywhere. Then the 2008 crash hit, followed by another glut in the mid-2010s. Orders evaporated. OEMs shut factories, laid off skilled labor, and merged their supplier bases. The “painful overbuild” became a corporate cautionary tale as the turbine market crashed in 2018.

By the late 2010s, renewables dominated investor attention. Gas turbine divisions were treated as cash cows, not growth engines. Capacity stayed flat even as demand for electricity quietly began to climb again. Then came AI.

Starting in 2023, data-center power requirements exploded. Hyperscalers that once drew tens of megawatts suddenly needed hundreds. Natural gas looked like the fastest, most practical bridge to new capacity, but the world only had three major manufacturers left, and none had invested to scale. With the fresh memory of 2018 and some uncertainty about how much energy the AI buildout would actually require, they weren’t rushing to ramp capacity again.

Siemens Energy now reports a €136 billion backlog, the largest in its history. GE Vernova has roughly 55 gigawatts of gas orders in its queue and plans to expand from about fifty to eighty heavy-duty units a year by 2026. Mitsubishi Power says it will double production, but is already sold out into 2028. Even if all three deliver on their expansion plans, total output might rise only twenty to twenty-five percent—nowhere near enough to meet demand.

Anatomy of a Shortage

Every layer of the supply chain is fragile. The high-temperature blades and vanes inside the turbine are cast from exotic nickel alloys by a few companies such as Howmet Aerospace ($76B market cap) and Precision Castparts (acquired by Berkshire Hathaway for $37B). The massive forged rotors that hold everything together are produced by a handful of plants worldwide, including Japan Steel Works. The heat-recovery steam generators that make combined-cycle plants efficient are backlogged. Transformers, switchgear, and control systems are just as scarce.

The result is a perfect storm of concentration and pricing power. One energy CFO told us that total project capex costs have doubled in just fifteen months, from roughly $1,000/kWh to more than $2,000/kWh. “It’s cartel pricing,” he said. “There’s nowhere else to go.” Reuters now cites installed combined cycle gas turbine costs moving from ~$1,000/kW to $2,000–$2,500/kW on recent projects, driven by turbine scarcity, HRSG/transformer backlogs, and EPC/labor constraints. Some developers are scouring Alberta and the Gulf Coast for refurbished turbines from the 1990s because new ones are unavailable. Others, including major industrial conglomerates, are buying units on spec simply to guarantee they have equipment for future plants. Koch Industries is rumored to be doing the same.

Implications for the AI Buildout

Natural gas was supposed to be the fast path. Today, its build timelines look a lot like those of emerging nuclear small-modular-reactor projects. It’s no surprise that Microsoft, Google, and others are signing enormous power-purchase agreements with nuclear and geothermal developers such as Oklo, Kairos, TerraPower, and Fervo.

A chief strategy officer at one hyperscaler told us he’s signing these LOIs freely, while assuming most will never reach operation. Even among those that do, he expects only a fraction of the current terms to survive to commissioning. But that hardly matters. For hyperscalers, the act of signing unlocks investment flows, de-risks sites, and signals seriousness to regulators. It’s a rational strategy in an irrationally tight market: spray capital at every potential source of electrons, then back the few that deliver.

Our Perspective

The turbine shortage is not just an isolated equipment problem. It exposes the fragility of the entire industrial stack: the transformers, EPC labor, HRSGs, and switchgear that all depend on long, brittle supply chains. Turbines just happen to be where the pressure shows first.

For startups, there are enormous opportunities across this landscape. Predictive-maintenance networks for turbine fleets. Software that helps EPCs and utilities coordinate procurement. Financing models that let developers pre-buy critical equipment. Materials or manufacturing breakthroughs that reduce lead times. Even the idea of a fuel-agnostic turbine that runs on gas today and integrates with a nuclear reactor tomorrow feels less like science fiction than a commercial inevitability.

Still, solving this likely won’t be as simple as building another “Anduril for turbines” (though we’d entertain it). Heavy industrial capacity takes years, patient capital, and government partnership to stand up. The payoff is profound. Whoever helps unjam these bottlenecks will sit at the center of the next trillion-dollar buildout and power the world’s compute.

We’ll continue exploring these supply-chain chokepoints in future issues, from turbines to transformers to the EPC networks that knit them together. As always, we’re early in our journey into energy and learning in public. If you’re building, financing, or operating anywhere in this bottleneck, whether you make turbines, HRSGs, or just want to power the next wave of AI, we’d love to hear from you.