# Catalyst: Turbines Are Sold Out to 2032, and the Big Three Still Won't Build More Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/catalyst-2026-10-29-the-new-landscape-for-gas-turbines/ > Notes on the 29 October 2026 episode of Catalyst with Shayle Kann, where Kann and gas turbine market analyst Tony Brough take stock of the order book: 2026 global orders now tracking 126 to 132 GW, delivery slots at the big three pushed out to 2030 through 2032, and 20 to 35 MW units quoted at $1,900 to $2,500 per kilowatt. Educational listening notes only — no investment advice, no price targets, and the judgements here may be wrong. Published: 2026-10-29 Locale: en Tags: gas turbines, podcast-notes, data centres, power shortage, chokepoints ![Afternoon in an aircraft boneyard, a jet engine freshly unbolted from a wing being lowered by crane onto a flatbed trailer, the engine lit by direct sun, rows of retired airliners receding one behind another into the heat haze](/covers/catalyst-2026-10-29-the-new-landscape-for-gas-turbines-cover.png) > Population, when unchecked, increases in a geometrical ratio. Subsistence increases only in an arithmetical ratio. > > —— Thomas Malthus, "An Essay on the Principle of Population" (1798) In the 29 October 2026 episode of Catalyst with Shayle Kann, the host brings back gas turbine market analyst Tony Brough, president of Dora Partners and Company, to walk through the order book. Brough says that a year ago he forecast roughly 101 GW of global orders for 2026; he is now tracking 126 to 132 GW. Reserve a slot with one of the big three today and delivery lands in 2030, 2031 or even 2032, with some product lines not taking another order for five years. His conclusion is that the gap gets filled by a crowd of non-traditional solutions, most of which will not survive the cycle. That conclusion rests on demand turning somewhere around 2028, and he concedes he undershot this year's order number by more than twenty percent. ## Sold out to 2032, and still not building more Start with the numbers. GE Vernova is sitting on about 53 GW of firm orders plus roughly 63 GW of slot reservations — money transferred to hold a place in the queue, not yet converted into firm orders. Siemens Energy's backlog runs about 70 GW, Mitsubishi Power about 35 GW. GE Vernova alone comes to nearly 120 GW, roughly the other two combined. On the capacity side, by 2030 GE Vernova goes from about 20 GW a year to 30, Siemens Energy from 16 to 23, and Mitsubishi Power close to doubles. Each of the three lands between 23 and 30 GW a year by 2030. Everyone else — Baker Hughes doubling, Solar Turbines close to doubling, Doosan entering — tops out at 5 GW or less. Then demand. Brough puts North American data centre capacity demand at 150 to 160 GW by 2030. The big three together will have something over 70 GW a year by 2030, and that is global capacity, the same lines feeding grids, industry and the rest of the world. So why not chase the demand? Brough says the makers are still guarding against an AI bubble bursting and will stay disciplined on capital spending. That sounds conservative until you look at pricing: units in the 20 to 35 MW class now go for $1,900 to $2,500 per kilowatt, against roughly $400 to $500 a few years ago. He expects prices to settle around $800 to $900 per kilowatt, still double what that class fetched five years back. And there is a first in the industry's history buried here: turbine makers historically earned their money in the aftermarket and struggled on new units, and now the largest ones make more profit on new units than on service. Expanding capacity until the price falls back to five hundred dollars would dismantle that with their own hands. For a sense of scale: the units data centres buy mostly sit between 7 MW and 100 MW, the popular class is 14 to 17 MW, with a step up at 25 to 35 MW for aeroderivatives — aircraft engines adapted to generate power on the ground. The 300 MW-plus machines Elon Musk ordered are the exception. Who fills the gap? Brough sorts the entrants into four groups and has reservations about three of them. The smallest are reciprocating engines under 3 MW, scaled-up versions of the generators you see on building sites and at hospitals, from firms like Cummins and Generac. He says some sites are planned with five to eight hundred of these units each, a number he called mind-boggling. His read: the moment demand turns, those orders go first, because wiring eight hundred units to run in sync and then maintaining each one is a logistics and operations nightmare. Second are the startups designing turbines from a clean sheet. Brough quotes a boss from his General Electric days who called a gas turbine "a symphony in metal." Designing a sellable turbine from the ground up having never done it, he says, could be a ten to twenty year undertaking — and one of these two companies has already lost its largest order. Supercritical CO2 cycles get a different verdict: that approach leans on steam turbine technology, the technical leap is far smaller, and he treats it as a workable alternative, if one with limited effect on market size. The third group he likes: converting retired jet airliner engines into stationary generators. FTAI is the best known, there is also Nova Power, and he knows of at least one to three more not yet public. He is specific about the engineering: you have to deal with the big fan at the front built to make thrust and get rotating torque out instead, and you have to redo the combustion system, because nobody wants to burn liquid fuel at a power plant. Solve those two and it works, and tens of thousands of engines will come off the flying fleet over the next five to ten years. ProEnergy goes further, keeping only the external carcass and re-engineering every internal component in house. The fourth group is the genuinely new: Mainspring's linear generators and Bloom's fuel cells. His reservation matches the one about small engines — Mainspring strings ten units together for 2.5 MW, and that set wants about 6,000 square feet of ground, where the same footprint could hold 30 to 100 MW of gas turbines. ## But he undershot this market last year too The strongest counter in the episode comes from Brough himself. A year ago he forecast about 101 GW of orders for 2026; the number ran to 126 to 132 GW, close to the upside case he had drawn. The conservative side has already been wrong once, and his "starts shrinking after 2028" call carries risk on the same side. The second tension is also in the episode. He argues the long-term trend bends down, because annual solar additions run an order of magnitude above gas turbine additions and renewables are the cheapest source of electricity globally, the United States included. Yet both things hold in 2026: renewables are cheapest, and gas orders are at a record. Which tells you what this round of gas demand is buying — interconnection timing and dispatchability, with cost ranked below them. As long as that is scarce, the cheapest source will not hold gas orders down. He adds the admission himself: even in his worst case, order volumes ten years out exceed 2023. Third, on the small-engine logistics nightmare, Shayle Kann raised a counterexample on the spot that I think holds up better than Brough's answer. Behind the meter power — generation built on the data centre campus that does not wait for the utility grid — derives its entire value from arriving early. A fleet of small units makes N+1 and N+2 redundancy easier, and more to the point they are available now, where the big three are booked past 2030. Brough's reply was that given the current state of the market Kann is right, but these are the first things to stop being bought when demand turns. That answer treats bridge power as a one-off transition, and how long the transition lasts depends on how long the interconnection queue is — a number nobody in the episode puts forward. Fourth, the supply story behind the aeroderivative line bites his own price forecast. If tens of thousands of engines really are convertible, and the conversion hurdle is only the fan and the combustion system, then new supply arriving around 2030 is more than the few tens of GW the big three are adding. The $800 to $900 per kilowatt floor could just as easily be a waypoint. Fifth, a question about the ruler he measures with. He uses ground footprint against Mainspring and unit count against small engines, and both of those measure steady-state operating efficiency. What this round competes on is time to power, and land costs far less than a year of lost revenue from a late start. Judging a solution by power density per square foot assumes land is scarcer than time. Out on the plains of Texas and the American Midwest, land for a data centre has never been the scarce thing. ## My read: it turns on how long the interconnection queue is Start with a doubt you may have: the news says shortages out to 2032 and prices up fourfold in five years — can an industry like that be believed? Fair question, because every shortage story looks the same, and shortage stories end in only two ways: demand retreats, or supply catches up. This episode gives enough material to tell them apart, and the test does not sit with the turbines. My own judgement is that the decisive variable here is grid interconnection wait time, with turbine capacity second. The reason is that in three of Brough's four groups, survival hangs on the same condition. Small engines are expensive, numerous and hard to maintain, and they can be installed now. Aeroderivatives need a new fan arrangement and combustion system, and the engines are already sitting on the tarmac. Clean-sheet turbines take ten to twenty years, so for this round they do not exist. All of them sell the same thing: time. So here is the dividing line. If interconnection waits shorten — permitting speeds up, transmission gets built, utilities free up capacity — the premium on arriving early disappears, Brough's "first orders to be cut" holds, and a retreat to $800 to $900 per kilowatt follows naturally. If the queue keeps stretching past 2030, the operating nightmare of eight hundred small engines is a cost worth bearing, because the alternative is a whole data centre sitting idle. That also explains a familiar puzzle: why industry news can be uniformly good while the related stocks run for a while and then start chopping. A sold-out order book is a fact that has already happened; what gets traded is the stretch after 2028, and on that stretch even someone who has followed this market his whole career will only say it shrinks without saying by how much. The distance between his worst case — orders ten years out still above 2023 — and his central case is wide enough to support two opposing valuations. Looking at chokepoint questions like this, I have a habit: ask first whether the scarce layer has a grip on pricing. This episode gives a clean answer. Making more profit on new units than on service has never happened in this industry, which says the big three are collecting scarcity rent, and that they are choosing not to spend capacity expansion eating that rent. Giving up share to defend price is behavioural evidence of pricing power, harder than any adjective about tight supply. Look the other way at the suppliers of those eight hundred small engines: they take the orders leaking out of the same gap, none has capacity above 5 GW, and none sets the price. ## Worth a look - The 29 October 2026 episode of Catalyst with Shayle Kann, the source for this piece; the show is produced by Latitude Media - The same guest's June 2025 conversation on turbines, which this episode refers back to repeatedly - Public filings and investor materials from GE Vernova, Siemens Energy and Mitsubishi Power, where the backlog and capacity plans can be checked directly - Public technical materials from FTAI, ProEnergy, Mainspring and Bloom Energy - Interconnection queue data published by the regional grid operators in the United States ## The one thing to take with you **Pricing power comes from being willing to walk away from a sale.** What stayed with me is the pairing: the big three are sold out to 2032, and they are still only taking capacity from 20 GW a year to 30. They can do the arithmetic on what expansion does to the quote — back down to five hundred dollars — so they would rather let the gap leak out to somebody else. And the companies catching that leak have full order books and no say over price. Here is something I have tried that works outside investing too. Take the last thing somebody asked you for — extra hours, a favour, borrowing something, an hour of listening — and write down what their alternative would be if you said no, and how long it would take them to find it. The more expensive the alternative, the more your position looks like the big three; if they can swap you out in three minutes, you are one of the eight hundred small units. Log three of these over the next month, then look at the times you said yes and count how many of them the other person did not need you for.