Constellation Energy: The Nuclear-Powered AI Winner
Constellation Energy (NASDAQ: CEG) is the largest operator of nuclear power plants in the United States, and it has become one of the most-searched names among investors trying to understand how the AI data center boom gets its electricity. The company owns more than 20 nuclear reactors and generates the biggest slice of carbon-free power in the country, which is exactly the product AI builders now want under long-term contracts. Its role in the AI story is simple to state: when a hyperscaler signs a 20-year deal for firm, around-the-clock power, Constellation is one of the few companies that can deliver it at scale.
The reason Constellation Energy keeps coming up in AI conversations is the math problem Marc Chaikin spends much of his “The End of Elon?” presentation on. Chaikin, the 60-year Wall Street veteran behind the Power Gauge Report, walks through the electricity squeeze with real numbers: he cites IEA projections that data center electricity use could roughly double by 2030, a Bloomberg study finding electric bills up “by an average of 267% in areas with high data-center concentrations,” and Elon Musk’s Colossus campus covering “1,156 acres” while drawing about two gigawatts. Whether or not you accept every figure, the direction is not controversial. AI runs on power, and the power has to come from somewhere.
Why nuclear is the AI story Constellation fits
Nuclear has three properties the AI industry needs. It is firm, meaning it runs day and night regardless of weather. It is carbon-free, which matters to tech companies with public climate commitments. And it is already built, which matters when grid interconnection queues stretch for years. That is why the marquee deal in this space involves Constellation directly: the company announced a 20-year agreement to supply Microsoft with power by restarting a reactor at Three Mile Island, the site it already operates. Restarting an existing plant is faster and cheaper than building anything new, and Constellation has more restart-ready capacity than anyone.
The market has already rewarded the theme. In a separate Stansberry resource presentation we reviewed, analyst Whitney Tilson cites Constellation Energy among the stocks that ran as “AI is colliding with” new sectors, with the stock up around 149% at one point. Chaikin’s own track-record reel in the Atlas presentation includes Vistra (VST), a rival power generator serving data centers, which he says his Power Gauge flagged before a run of more than 1,300%.
Where Constellation fits in the American Atlas thesis
Here is the honest caveat: Constellation Energy is not one of the stocks Chaikin reveals in the American Atlas pitch. His free pick in that presentation is AMD, the chipmaker whose Instinct accelerators power Department of Energy supercomputers like Frontier and El Capitan. His broader argument, which we cover in our full review of the End of Elon presentation, is that a new class of FP64-precision “AI micro clusters” at Oak Ridge National Laboratory could shift what AI compute looks like.
The energy angle is where the two stories connect. Chaikin’s claim is that a micro cluster like Lux, the first FP64 machine at Oak Ridge, draws “as little as 24 megawatts,” roughly one percent of Colossus’s appetite, while doing scientific work that today’s massive campuses cannot. If he is even partly right about that efficiency gap, the marginal demand for grid power from AI shifts too: fewer hyperscale campuses drawing two gigawatts, more distributed facilities that existing nuclear plants could serve directly. Constellation’s advantage in either scenario is the same. More AI of any flavor means more electricity, and nuclear is the premium supplier.
For investors evaluating the Atlas opportunity specifically, Chaikin’s report portfolio targets 17 companies that “have signed deals” to work on the buildout, which he says include names “absolutely tiny compared to AMD.” Constellation would be the opposite profile: a large, liquid, dividend-paying utility. The two are complementary exposures rather than substitutes.
What to consider
Constellation’s AI upside is already partially priced. The stock re-rated sharply once the Microsoft deal put nuclear-at-AI on the front page, and its valuation now assumes the data center buildout continues. Rates matter too: utilities are capital-hungry, and higher-for-longer borrowing costs press on expansion plans. Regulated markets cap some of Constellation’s pricing power, and reactor restarts, however promising, still face Nuclear Regulatory Commission schedules. None of these are reasons to dismiss the thesis; they are the factors that determine whether an AI-power investment compounds or stalls.
The bull case beyond AI is real as well. The federal government has leaned hard into nuclear as part of national energy security, and Constellation’s fleet positions it for everything from defense-facility contracts to hydrogen production. In that sense, the AI boom is an accelerant on a business that was already durable.
The bigger picture
Constellation Energy is the cleanest large-cap expression of a simple fact: AI is an electricity business as much as a software one. Chaikin’s American Atlas presentation is about where the computing happens; Constellation is about what powers it. Both are worth understanding before the next leg of the buildout makes the front page.
If you want to go deeper on the Atlas side of the story, start with our full End of Elon? American Atlas review, then read how the AI data center energy problem frames the whole pitch and the 17-company American Atlas universe Chaikin teases. For the machines themselves, see our piece on Oak Ridge, Frontier, and the Lux supercomputer. And if you want Chaikin’s portfolio report details, the American Atlas Portfolio is explained here. If you want to hear Chaikin’s argument in his own words, watch the presentation here.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.