Data Centers Are Fueling a Tiny Nuclear Hype Cycle

Investment in smaller, next-generation nuclear reactor designs is booming, with a flood of capital pouring into scaled-down models known as small modular reactors — or, if they’re extra tiny, microreactors. In just the past few weeks, Valar Atomics announced a $1 billion Series B, while Antares Nuclear closed its $470 million Series C. The two companies are attempting to serve different customers — Valar is targeting hyperscale data centers, while Antares is building for off-grid military applications — but both are betting on the same premise: that smaller, factory-built reactors can deliver reliable, carbon-free power far more quickly, flexibly, and cheaply than traditional large-scale nuclear plants.

Venture capital is eating it up. In addition to Valar and Antares’ raises this year, SMR startup X-Energy went public in April, raising over $1 billion at a $9.1 billion valuation. Last year alone, SMR companies TerraPower, Last Energy, Radiant Industries, Aalo Atomics, Arc Clean Technology, and Stellaria all raised rounds.

It seems like every week brings another announcement about an SMR company hitting a new milestone or a microreactor raising a new round. But some industry experts aren’t buying the hype. One 2024 report by the Institute for Energy Economics and Financial Analysis summarizes it neatly with the title, “Small Modular Reactors: Still too expensive, too slow and too risky.” One of the report’s co-authors, Dennis Wamsted, thinks this blunt analysis has held up remarkably well.

“I still think that’s one of the best-titled reports we ever wrote,” he told me, arguing that nothing in the past two years has changed his fundamental analysis of the sector. “I think it’s just as overhyped as it was a few years ago. There is a shiny new object mentality to SMRs. They’re going to work perfectly right out of the box.” Instead, the report argues, borrowing a phrase from NextEra Energy CEO John Ketchum, SMRs are “an opportunity to lose money in smaller batches.”

The report came out about six months after NuScale — still the only SMR company with a design certified by the U.S. Nuclear Regulatory Commission — canceled its inaugural project in Idaho before construction even began. It’s been a wild ride ever since: Buoyed by investor excitement over an artificial intelligence-driven nuclear renaissance, NuScale’s stock soared last year before losing most of its value once again as the company posted major losses.

The AI boom has driven much of the surge in SMR interest, as hyperscalers scramble to procure power for a rapidly expanding fleet of new data centers. Google, Amazon, and Meta have signed agreements with SMR developers Kairos Power, X-energy, and TerraPower and Oklo, respectively. At the same time, bipartisan support for nuclear is growing. Recent Gallup polls show that 46% of Americans believe the U.S. should put a greater emphasis on nuclear power and that 61% support the technology overall. Other surveys suggest SMRs in particular enjoy even higher levels of favorability.

The Trump administration has gone all in too, signing executive orders directing the Department of Energy and Department of Defense to prioritize deploying small reactors at domestic military bases and spinning up the Reactor Pilot Program to expedite testing of 11 new advanced reactor designs outside the jurisdiction of the Nuclear Regulatory Commission. The program aimed to have three reach criticality — the point at which a nuclear reaction becomes self-sustaining — by this July 4th. Four microreactor companies ended up beating the deadline: Antares, Valar, Deployable Energy, and Aalo Atomics, while the Sam Altman-backed SMR company Oklo achieved criticality last week.

“Say I was an advisor to the Department of Energy,” Wamsted’s co-auther David Schlissel, formerly director of resource planning analysis at the Institute for Energy Economics and Financial Analysis, posited to me. “Even with the risk, the smart way to go is, let’s pick two or three designs and go out and build them. Build one of each. See which ones work and which ones don’t. But what’s happening is the exact opposite of that.”

Whether federal policy is creating a durable new industry or not, there are still plenty of situations where customers need clean, firm power and today’s options fall short. Solar-plus-storage is broadly useful, but matching nuclear’s 24/7 availability can require significant overbuilding. And when it comes to large-scale nuclear, a customer may need power sooner than when a project that big could feasibly come online.

Many customers are also simply unwilling to take on the risk of a multibillion-dollar, decade-long nuclear megaproject, which tend to run over time and budget. The only new reactors built in the U.S. since the Three Mile Island accident in 1979 — two huge Westinghouse AP1000 units capable of generating 1.1 gigawatts of power apiece — have become poster children for this risk. Units 3 and 4 at the Vogtle Electricity Generating Plant in Georgia came online in 2023 and 2024, respectively, roughly seven years late and tens of billions of dollars over budget. Georgia Power customers will be paying off Vogtle well into the 2050s.

This has left many SMR entrepreneurs and industry boosters convinced there simply must be a better way. "The only customers capable of buying a reactor that large are either nation-state governments or essentially state-backed utilities,” Jordan Bramble, Antares’ co-founder and CEO, told me.

In part because of this, Bramble rejects the idea that small reactors are even competing with large-scale nuclear in the first place, explaining that the either/or framing overlooks the fact that these designs attract distinct pools of capital. “What a venture capitalist in private equity is going to invest in versus a municipal bond investor or a utility investor is going to invest in are two totally different things,” he told me.

And while SMRs may eventually seek institutional capital too, Bramble points to recent funding rounds by Anthropic, OpenAI, and Commonwealth Fusion Systems as evidence of just how much money companies can attract in today's private market even before their tech has come down the cost curve. “I think when the upside equation is there, there’s near limitless money in venture and growth equity right now,” he told me.

True? Largely. Indicative of a bubble? Possibly.

One lesson many developers took from NuScale seems to be about customer selection. While NuScale intended to serve a coalition of small, price-sensitive municipal utilities, today’s SMR startups are targeting early adopters with more room in their budgets: AI hyperscalers, of course, but also military and defense customers and industrial companies such as chemicals and metals producers that can put both nuclear’s heat and electricity to use. Modular, factory-based production is central to many of their strategies, along with even smaller reactor designs. While NuScale sought to build 77-megawatt reactors, Valar is targeting 5 megawatts while Antares is building in the 100-kilowatt to 1-megawatt range.

But utility analyst Bill Tilles argues that scaling down further isn’t the answer. The fundamental issue with SMRs, he told me, is that they suffer from a "reverse economy of scale." That is, shrink the size of the reactor and the cost per watt of electricity produced goes up, not down. Add in a market crowded with dozens of these companies pursuing different reactor designs and fuel types but chasing the same data center, defense, and industrial customers, and it becomes difficult to see how any single one can attract the critical mass of customers needed to scale up a manufacturing line and become relatively cost-effective.

Of course, every SMR company says it’s uniquely positioned to emerge as a winner in what even Bramble acknowledges is an overcrowded field likely to see consolidation in the coming years through either mergers and acquisitions or outright failures. Still, he’s feeling confident in Antares’ decision to pursue the Department of Defense as a beachhead customer: In April, the Air Force selected the company to build a 500-kilowatt microreactor at a military base in San Antonio, set to come online in 2028.

“[Nuclear] actually was always a defense-first technology that eventually became commercial, and that’s how rocket propulsion worked. It’s how GPS worked. It’s how semiconductors worked. It’s even how the internet developed,” he told me. Bramble said he thinks Antares can follow a similar trajectory, riding the cost curve down before eventually bringing a grid-scale product to market.

While SMR skeptics may not be convinced this grid-scale goal is truly feasible, many do acknowledge that remote military bases offer a compelling, if niche, market for SMRs and microreactors. The military has operated nuclear-powered submarines for decades, so the concept of using small reactors in situations where conventional refueling is costly and dangerous is not without precedent. “You have these unique, price insensitive buyers that the government will try to encourage,” Tilles told me of remote deployments. “But one should not confuse that with anything resembling a commercial technology.”

That may be where the real debate lies — whether there are enough price insensitive customers for multiple companies to commercialize small reactors at scale and drive costs down.

There’s also the question of what the market will look like by the time these companies are ready to scale production — a milestone experts peg around the mid-2030s. Ultra-long-duration energy storage company Form Energy and advanced geothermal developer Fervo are already building out and turning on their first commercial projects, while multiple fusion companies are similarly targeting the mid-2030s for commercialization. If any or all of these technologies take off, they could reshape the market for clean, firm power — and thus the options available to SMRs’ potential customers.

But Benton Arnett, senior director at the industry group Nuclear Energy Institute, argues that multi-billion-dollar energy customers would be unwise to put all their eggs in one technological basket, betting that ultra-long duration storage or fusion alone will meet all their future energy needs. “You’ve got to have a diversity of investments and a diversity of plays so you can capture what’s going to be most available over the next 10 years, which can be really hard to predict,” he told me. He’s obviously betting SMRs will be among those technologies of the future. “I think everyone’s building right now not based on hype, but based on real dollars that are changing hands, building out this kind of new data center ecosystem.”

Bramble, for his part, thinks the hype cycle might be real. He just doesn’t see the exuberance as a negative for Antares or the industry at large. “Some of the most generational, economically transformational companies get built during a hype cycle,” he told me. “That was true of Google and Amazon in the dot-com bubble. This was true of the railroads. The best ones emerged during a period of mass overbuilding and overinvestment.”

So the question may not be whether the SMR boom will produce any winners, but how many — and how much capital investors and startups will burn in the process. Because while the Google of small nuclear may still be waiting to emerge, history suggests there will be plenty of nuclear equivalents of Pets.coms, Kozmo.coms, and Webvans along the way.