Doomberg Says the GE Vernova Trade Is "Done" and the Next Gas-to-Power Winner Is Whoever Escapes the Turbine Bottleneck
Anonymous energy commentators outline their views on gas, nuclear and AI power on the AI Investor podcast, October 15, 2026
Doomberg, the anonymous Substack publication run by former industry executives and commodity veterans, argued on the AI Investor podcast hosted by 247 Wall St that the easy money in gas-fired AI power infrastructure has already been made in turbine makers. "I think the GE Vernova trade, if you pushed us, is done. Everybody kind of knows that story. What's your edge?" The team said it does not make individual stock recommendations, but that the next leg of value creation lies in technologies that convert natural gas to electricity without sitting on the critical path of turbine production.
Turbine Shortage as a Setup for a Glut
Doomberg framed the problem not as a shortage of turbines but as a shortage of machines that can convert cheap gas into power cheaply enough to capture the discount gas offers. It invoked what it calls Taleb's law, the idea that "not all gluts lead to shortages, but I've never seen a shortage that doesn't lead to a glut." On that logic, GE Vernova's reported backlog of roughly 116 gigawatts, with orders now being taken for 2031, is a signal that the market is crying out for alternatives, and that those alternatives will eventually erode the scarcity premium.
Bloom Energy, whose solid oxide fuel cells are also sold out, was cited as evidence of that demand. Doomberg was candid that it has "mixed analysis" of the company and that "there's a lot of questions swirling around it," but treated Bloom as a market signal rather than a pick. The team said that if investing privately, it would look for a startup that lets data centers turn gas into compute without relying on turbine supply. "Closing of such arbitrage is where generational wealth is created." It also noted that Elon Musk has been making investments in this area.
Asked about Solaris Energy, which rents gas turbines to data center customers including SpaceX, the team said it was hearing of the company for the first time. It called the model a smart way to capture the bid-ask spread in turbine demand and expected a "boomlet" of revenue while speed remains the rate-limiting factor. The warning was blunt: the business model "is one technology invention away from being made obsolete," namely a cheap, quickly deployable device that turns gas into electricity. Doomberg said the key question for shareholders is capital allocation, and that returning cash through dividends or buybacks matters more than any hope of a durable moat. Host Austin compared the setup to neo-clouds that benefit from a temporary pricing arbitrage in GPUs.
Cheap Gas Is a Byproduct of Expensive Oil
The most counterintuitive point was why US natural gas trades near $3 per million BTU, roughly equivalent to $18 oil on an energy basis, while Brent is above $100 and diesel is tight following this year's war with Iran. Doomberg's answer is co-production. Oil and gas come up together from the same shale wells, so high crude prices push drillers to produce more, and the associated gas is given away or sold at rock-bottom prices where pipeline capacity is lacking. "The war in Iran has made natural gas cheaper than it otherwise would be simply because of co-production." Gas in the Permian Basin has at times traded at negative prices over the past 18 months, and Alberta and British Columbia face similar egress constraints.
The team expects this arbitrage to close through "engine switching," with US data center construction as one of the main outlets. Doomberg said its 2024 call that data centers would move off-grid and onto gas fields has been directionally right, citing the Chevron and Microsoft 2.7 gigawatt Permian project and SoftBank's 9 gigawatt gas build-out in Ohio. It admitted the timing was off. "A common theme is that we're just too early. Just because it should be that way doesn't mean it is." Early movers such as EQT in Appalachia were cited as exceptions.
Doomberg also pointed to cooling as a driver of siting, given that roughly half of a data center's energy goes to cooling. That favors Alberta, British Columbia and Iceland over the hot Permian. Alberta was called "the Texas of Canada," with Premier Danielle Smith working with the new prime minister to cut regulatory delay.
Rebutting the Gas Shortage Thesis
The team directly challenged the view, put forward by Matt Smith on another podcast, that gas demand will exceed supply by 2028 and trigger a crisis by 2030 as LNG exports and AI demand collide, potentially lifting gas from about 10% to as much as 40% of hyperscaler costs. Doomberg said Smith's supply forecasts are in line with its own, and well above those of "peak shale gas" pessimists. The disagreement is about policy: Smith assumes LNG exports are prioritized, while Doomberg believes Washington would toggle exports down because AI is a bipartisan national priority.
The second counterargument is supply response. Doomberg said a hypothetical $5 price floor would trigger "the utter stampede of drilling and new supply" that would shock observers. It put the break-even cost for leading Appalachian producers such as EQT at about $2 per million BTU, roughly equivalent to $12 oil, and noted that Appalachia alone would be the world's third-largest gas producer behind the US and Russia if it were a country. Its broader stance is that analysts consistently underestimate the technical capability of the oil and gas industry. "The size of the hydrocarbon resource base underground is best modeled as being infinite. Period."
Nuclear: Optimism Turned to Cynicism
Doomberg said it has gone "full cycle" on nuclear. It initially predicted correctly that AI demand would erode political barriers, but it has since become cynical as the sector filled with "fly-by-night hucksters and pump and dumpsters." Its core position is that no new technology is needed. Building 100 AP1000 or CANDU reactors on proven designs could produce units at $1 billion to $2 billion per gigawatt, as China does, instead of a multiple of that in the West. It said venture capital has wrongly convinced itself that fancy technology is the missing piece.
The team was dismissive of fusion, calling it "an unneeded technology that solves problems that don't exist," and of small modular and thorium designs, describing claims of less waste or no meltdown risk as "fantasies." France's fleet, built 50 years ago, still supplies two-thirds of its electricity. "Find me the technology barrier. There is no technology barrier. It's political." On waste, Doomberg said, "Nuclear waste is literally the greatest job of branding that opponents to a technology have ever done," noting that 95% of the embodied energy in uranium remains after use. The conclusion for investors is that nuclear will miss this window: "Natural gas is going to win."
Why the Grid Will Not Serve AI
On whether the grid can catch up with AI demand, Doomberg argued it should not try. The grid is a public good designed to underpin living standards, delivering reliable power at below marginal cost so that people consume more of it, and not a market meant to clear at the true marginal price. "You do not want to find out what the true clearing price of electricity is at the margin and then charge everyone in society that price." The team compared it to the national highway system, which should not be degraded for the benefit of one industry. The implication is that behind-the-meter generation is the structural answer, consistent with the host's observation that self-powered projects have risen from about a third to roughly 50% of the total in six months.
Where Doomberg Would Look: Midstream
Asked to name one energy sector to add to an AI-focused portfolio, Doomberg pointed to midstream, on which it published a premium deep dive last month covering 10 companies. The logic is that pipeline operators are paid on volume rather than commodity price, so they benefit if pipelines are needed to connect gas fields to data centers. The team described them as capital compounders with fortress balance sheets, some leverage and steady dividends of 8% to 12% a year. Names raised in the discussion included Enbridge, Enterprise Products Partners, Kinder Morgan, Energy Transfer and TC Energy, the last described as having near-monopoly control of gas movement in Alberta. EQT was cited as an integrated Appalachian producer that now owns the Mountain Valley Pipeline and is moving toward direct LNG sales.
Doomberg was explicit about the limits: "These aren't going to be your 10-baggers." Investors also face tax complexity, since many midstream names are master limited partnerships that issue K-1s rather than 1099s.
Geopolitical Framing and the Bubble Question
The conversation opened with a strongly stated geopolitical view. Doomberg argued that a "World War III" began in 2014 as a battle over US dollar primacy, and that the US and its allies are losing. It said NATO has been defeated in Ukraine and that President Trump "was defeated by Iran," first in June 2025 and again in February and March of this year, with ominous implications for the midterms. It described the US-China AI race as one fueled by gas on one side and coal on the other. These are opinions presented as a mental model, not analysis grounded in the energy data, and investors should weigh them accordingly.
On markets, Doomberg called AI a bubble that is also transformative. "It is simultaneously a very large bubble and an incredibly transformative technological leap. Both things are true." It compared the situation to shale, which destroyed equity while reshaping geopolitics. It believes demand for inference is effectively infinite, but said investors should sell into strength, and should know whether they are "renting the stock or owning it."
AI Will Be Disruptive for Incumbents
Doomberg said new AI-native businesses, not incumbents, will capture the value. "AI is going to expose just how many useless white-collar jobs there are," and selling AI to existing companies will not work. It expects well-capitalized entrants, and frontier AI labs with spare investor capital, to buy channels in boring industries and rebuild them around AI, private equity roll-up style. Its contrarian conclusion is that "the value of authentically human is going to skyrocket," favoring authentic human content and technologies that verify it. The team said it has gone dark on X and is concentrating on Substack, and that it is writing a book.