Splitting the Atom is Now the Easy Part: modular reactors for data centers

There’s a beige steel building going up near Interstate 40 in Oak Ridge, Tennessee. It doesn’t look like the future of nuclear power. It looks like the place where someone stores forklifts.

When it opens, X-energy says it will be able to produce around 700,000 uranium fuel spheres a year. That sounds like a lot until you translate it into reactors. It’s enough fuel for up to eleven Xe-100 reactors. Not one hundred. Not one thousand. Eleven.¹

That building may tell us more about when advanced nuclear power reaches American data centers than almost any reactor announcement. The reason is counterintuitive: the reactor may no longer be what’s holding nuclear back. Progress inside the reactor is beginning to outpace the value chain around it.

On July 4, 2026, Aalo Atomics brought a small test reactor to criticality in Idaho. “Criticality” sounds like something went wrong. It’s actually the point at which a reactor sustains a controlled nuclear chain reaction. Aalo was the fourth new American reactor to reach that point in about a month. The federal government hoped for three criticalities by Independence Day. The industry delivered four.²

That’s real progress. It suggests that advanced reactor teams can now design, assemble and operate test reactors much faster than most people assumed even a year ago. That said, it doesn’t mean fleets of commercial reactors are about to start powering data centers. The breakthroughs have made real progress, but in the process they've exposed a new set of far less glamorous (yet very real) bottlenecks: fuel processing, major electrical equipment and grid access. Those parts of the value chain can now set the pace.

The contrast is almost comical. Inside the reactor, engineers are controlling a self-sustaining nuclear chain reaction. Outside, the project may be waiting for uranium gas to become powder, for a transformer factory slot or for permission to connect to the grid. The industry spent decades learning how to split atoms. It may now wait on industrial chemistry, electrical steel and interconnection rights.

Said differently, the nuclear problem is becoming less nuclear. When a visible technical bottleneck begins to ease, value often migrates to the less visible parts of the system that remain scarce. Companies that can supply the missing processes, equipment or infrastructure may gain an important position in advanced nuclear without ever designing a reactor.

Follow the money

Hyperscalers are making serious commitments to advanced reactors. Google’s agreement with Kairos Power covers up to 500 megawatts, with the first project planned for 2030 and additional deployments through 2035. Amazon is backing X-energy deployment that could exceed 5 gigawatts by 2039. Meta’s January 2026 agreements with Vistra, Oklo and TerraPower cover as much as 6.6 gigawatts of new and existing nuclear capacity.³

These are serious commitments, but the advanced reactor portions are mostly bets on the 2030s. By the companies’ own schedules, they won’t do much to solve the immediate data center power shortage. The hyperscalers’ near-term behavior points somewhere else. Microsoft is supporting the restart of Three Mile Island Unit 1. Amazon expanded its agreement for power from the existing Susquehanna plant. Meta signed a twenty-year agreement tied to the Clinton plant. Google is working with NextEra Energy to restart Duane Arnold.⁴

Their public announcements emphasize advanced nuclear. Their near-term purchasing decisions emphasize existing reactors whose fuel supply, major electrical equipment and grid connections are already largely in place. That’s the difference between stated ambition and revealed preference. Press releases tell us what companies want to make possible. Their capital allocation tells us what they believe can actually deliver power.

In other words, the nuclear data center market is splitting in two. The first is a near-term race to secure power from existing plants, restarts and uprates. The second is a longer-term effort to build a new reactor industry and the value chain underneath it. The timing of that second market will depend on three practical questions: Is usable fuel available? Can the project obtain the major equipment it needs? Does it have a credible path to the grid?

The demonstrations routed around the commercial system

The four recent reactors were test units operating through Department of Energy programs rather than the normal commercial NRC pathway. That was enough to prove the reactors could operate. It didn’t prove that future fleets can be licensed, fueled and deployed through a commercial value chain at scale.⁵

That distinction matters because the demonstrations cleared the reactor hurdle by temporarily routing around parts of the commercial deployment system. They showed rapid progress inside the reactor. They didn’t show that the value chain outside it can yet support a fleet.

We can enrich the fuel. We can’t yet finish enough of it.

Most advanced American reactor designs require high-assay low-enriched uranium, usually called HALEU. Domestic enrichment is beginning to return, but enrichment isn’t the end of the process. It produces uranium hexafluoride gas. Most reactor manufacturers can’t use that gas directly. It must first be deconverted into metal or oxide, then fabricated into the particular fuel form a reactor requires.⁶

In other words, we’ve spent billions rebuilding the ability to make the right gas, only to discover that we still can’t turn enough of it into something a reactor can eat.

The Department of Energy has selected six companies to provide HALEU deconversion services. That’s progress, but an award isn’t the same thing as commercial capacity operating at scale. Fuel fabrication is moving too. X-energy’s TRISO-X plant in Oak Ridge is under construction and is expected to produce enough fuel for up to eleven Xe-100 reactors a year.⁷

Eleven reactors isn’t inherently too few. Reactor sizes, initial cores and refueling schedules differ. It does show how quickly multigigawatt ambitions run into the arithmetic of physical production. The reactor rendering can scale instantly. The fuel factory can’t.

Then there’s the steel box

A generator step-up transformer raises the voltage of electricity leaving a power plant so it can enter the transmission system. It isn’t specifically nuclear. A new reactor needs one, but so does a gas plant and almost every other large generating project.

That’s what makes this bottleneck difficult to escape. Switching reactor designs doesn’t remove it. Switching from nuclear to gas doesn’t necessarily remove it. Much of the new power system is competing for the same factories, materials, engineering capacity and skilled labor. Large generator transformers can carry lead times measured in years.⁸

The visual drama is almost absurd. Inside the reactor, engineers are controlling a chain reaction. Outside, the project may be waiting for a steel box. A reactor can work perfectly and still produce no commercial power because another supplier hasn’t delivered it.

Grid access may be the most valuable thing at the site

At the end of 2025, about 2,061 gigawatts of proposed generation and storage capacity remained in US interconnection queues. The total was lower than a year earlier partly because projects withdrew, not because they connected. Long timelines and high withdrawal rates remain structural barriers to new power plants.⁹

This helps explain the appeal of existing nuclear stations. They don’t just contain reactors. They occupy sites with transmission infrastructure and established paths to the grid. Those assets can take years to obtain.

A restart is still difficult. Existing equipment must be inspected, repaired or replaced. Regulators must approve the work. Operators and maintenance teams must be rebuilt. Existing plants simply begin with pieces of the value chain that new projects must still create. The hyperscalers aren’t only buying nuclear generation. They’re buying years of solved problems around the reactor.

Where the value is migrating

Our value chain analysis decomposed on-site micro reactor deployment for American data centers into more than 1,500 distinct technology and nontechnology factors across the fuel cycle, reactor, balance-of-plant equipment, licensing, site development, grid interconnection, operations and decommissioning. The counts aren’t weights, and they don’t prove that one category matters exactly twice as much as another. What they reveal is the breadth of ways a technically successful reactor can still fail to become a commercial power project.

That changes how we should read the next announcement. A reactor milestone tells us something important about the reactor. It doesn’t tell us that usable fuel is available, major equipment has been ordered or a path to the grid has been secured.

It also changes where investors, innovators and entrepreneurs should look. Companies don’t have to build reactors to participate in advanced nuclear. They may create more value by solving the scarce problems around them. A new deconversion process, a faster way to manufacture critical electrical equipment or a better path through interconnection can become strategically important precisely because those capabilities are less glamorous and less crowded than reactor design.

As bottlenecks migrate, value tends to migrate with them. Companies that secure scarce positions in fuel processing, equipment or grid access may gain pricing power, faster growth and greater industry influence while those constraints persist. That won’t last forever. Bottlenecks eventually attract capacity, substitutes and competition. The opportunity belongs to those who see the shift early enough to build or control the missing pieces before the rest of the market catches up.

The first wave of nuclear-powered data centers will therefore be shaped less by reactor design than by the strength of the value chain behind it. The useful question isn’t how many megawatts appear in the next headline. It’s whether the project has secured the fuel, major electrical equipment and grid access needed to deliver them. Until those parts of the value chain move, the timeline probably hasn’t moved much either.

The reactor may be the sexy part that gets the attention, but the value chain is the boring part that turns on the lights. It may also be where much of the value is now moving.

Endnotes

1. X-energy, “X-energy Begins Vertical Construction for First-in-the-Nation Advanced Nuclear Fuel Fabrication Facility,” November 17, 2025. https://x-energy.com/news/x-energy-begins-vertical-construction-for-first-in-the-nation-advanced-nuclear-fuel-fabrication-facility/

2. U.S. Department of Energy, “Department of Energy Celebrates First Advanced Reactor Criticality,” June 4, 2026; “Department of Energy Celebrates Second Advanced Reactor Achieving Criticality,” June 18, 2026; “U.S. Department of Energy Meets President Trump’s Goal, Delivers Third Advanced Reactor Criticality,” July 1, 2026; and “Department of Energy Celebrates Fourth Criticality Ahead of July 4th Goal,” July 2026. https://www.energy.gov/ne/us-department-energy-reactor-pilot-program

3. Google, “Google signs advanced nuclear clean energy agreement with Kairos Power,” October 14, 2024. https://blog.google/company-news/outreach-and-initiatives/sustainability/google-kairos-power-nuclear-energy-agreement/; Amazon, “Amazon signs agreements for innovative nuclear energy projects,” October 2024. https://www.aboutamazon.com/news/sustainability/amazon-nuclear-small-modular-reactor-net-carbon-zero; Meta, “Meta Announces Nuclear Energy Projects, Unlocking Up to 6.6 GW,” January 9, 2026. https://about.fb.com/news/2026/01/meta-nuclear-energy-projects-power-american-ai-leadership/

4. Constellation Energy, “Constellation to Launch Crane Clean Energy Center,” September 20, 2024. https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html; Talen Energy, “Talen Energy Expands Nuclear Energy Relationship with Amazon,” June 11, 2025. https://ir.talenenergy.com/news-releases/news-release-details/talen-energy-expands-nuclear-energy-relationship-amazon/; Constellation Energy, “Constellation, Meta Sign 20-Year Deal for Clean, Reliable Nuclear Energy in Illinois,” June 3, 2025. https://www.constellationenergy.com/newsroom/2025/constellation-meta-sign-20-year-deal-for-clean-reliable-nuclear-energy-in-illinois.html; NextEra Energy, “NextEra Energy and Google Announce New Collaboration to Accelerate Nuclear Energy Deployment in the U.S.,” October 27, 2025. https://newsroom.nexteraenergy.com/NextEra-Energy-and-Google-Announce-New-Collaboration-to-Accelerate-Nuclear-Energy-Deployment-in-the-U-S

5. U.S. Department of Energy, “U.S. Department of Energy Reactor Pilot Program.” https://www.energy.gov/ne/us-department-energy-reactor-pilot-program

6. U.S. Department of Energy, “HALEU Deconversion Services.” https://www.energy.gov/ne/haleu-deconversion-services; U.S. Department of Energy, “HALEU Availability Program.” https://www.energy.gov/ne/haleu-availability-program

7. U.S. Department of Energy, “Energy Department Awards Contracts for High-Assay Low-Enriched Uranium Deconversion Services,” 2024; X-energy, “X-energy Begins Vertical Construction for First-in-the-Nation Advanced Nuclear Fuel Fabrication Facility,” November 17, 2025.

8. Wood Mackenzie, “Mind the gap: tackling supply-chain challenges in the electric T&D sector,” October 2025. https://www.woodmac.com/news/opinion/mind-the-gap-tackling-supply-chain-challenges-in-the-electric-td-sector/; Reuters, “US power companies scramble to secure equipment as surging data center demand strains supplies,” July 9, 2026.

9. Lawrence Berkeley National Laboratory, “Queued Up: 2026 Edition,” July 1, 2026. https://emp.lbl.gov/queues