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

Something unusual happened in advanced nuclear this summer. Within about a month, four new American advanced reactors reached criticality. The Department of Energy had set a goal of getting at least three there by July 4. There were four.¹ That's a good thing, right?

Yes, but there are some important caveats. These were test or demonstration reactors, not commercial power plants, and the criticality demonstrations were conducted at zero power. They don’t tell us whether advanced reactors can be manufactured economically at scale, how quickly commercial plants can get licensed, or what they’ll ultimately cost to operate. Still, getting four different advanced reactor designs to criticality in about a month is real progress.¹

So what's next? If the reactors themselves are making progress, what are the next big ticket bottlenecks standing between where we are now and a world with more accessible (and safe) nuclear power?

To try and learn more about this, my team recently analyzed the value chain for on-site microreactors serving American data centers, breaking the system into more than 1,500 individual technology and nontechnology factors. We looked at the reactor, but also everything around it that has to work for that reactor to ultimately produce usable power.

Three major constraints rose to the top :

1. HALEU supply and deconversion. Many advanced reactors require high-assay low-enriched uranium, or HALEU. Commercial supplies of HALEU enrichment services in the United States remain limited, and DOE expects near-term requests for the HALEU it can provide to exceed its available supply before commercial HALEU becomes available. There is also limited capability for the next step of deconverting enriched HALEU into a form suitable for fuel fabrication.² ³

2. Major electrical equipment. Advanced reactors need some of the same large transformers and other electrical equipment that everyone else building power generation needs. Demand for this equipment has grown much faster than supply, and lead times for large transformers used at substations and generators can reach three to four years.⁴

3. Grid access. Even after you’ve built the reactor and everything works, you still need somewhere for the electricity to go. Getting new generation connected to the US transmission system can take years.⁵

None of these problems has much to do with the underlying science of splitting an atom.

1. HALEU supply and deconversion

Let’s start with the fuel.

Many advanced reactors require HALEU, which is enriched to between 5% and 19.75% uranium-235. The United States is investing heavily in new domestic production. In January, the Department of Energy awarded $900 million each to Centrus and General Matter to expand domestic HALEU enrichment capacity over the coming decade.²

That capacity will take time to build. Based on information available as of April 2026, DOE expected to have about 21.2 metric tons of uranium in HALEU available to advanced reactor demonstration companies by the end of 2028 and 23.4 metric tons by the end of 2030. DOE officials told the Government Accountability Office that demand would likely exceed what DOE could supply before commercial HALEU becomes available. How large that gap becomes will depend in part on how quickly the reactor projects themselves progress.³

Then there’s another step that’s easy to overlook. Enrichment produces HALEU as uranium hexafluoride. Before it can be fabricated into reactor fuel, it has to be deconverted into another form, typically an oxide or metal. DOE has contracted with six companies that can compete for HALEU deconversion work, but GAO found that this capability does not yet exist commercially for HALEU.³

Interestingly, fuel fabrication itself doesn’t appear to have the same problem. Industry stakeholders interviewed by GAO generally didn’t identify fabrication as a constraint. Some reactor developers are planning their own fabrication facilities or expect to use national laboratory facilities.³ So the issue is more specific than a general shortage of nuclear fuel. The constraints we found are concentrated around near-term HALEU availability and deconversion.

2. Major electrical equipment

This one surprised me a little more.

Once a power plant generates electricity, large transformers and other high-voltage equipment are needed to move that power into the electrical system. Advanced nuclear needs this equipment, but there isn’t a separate transformer industry waiting around for nuclear projects. They’re buying from the same supply base serving the rest of the electric power system.

That supply base is already under considerable strain. The Department of Energy reported this year that lead times for large transformers used at substations and generators have grown from roughly three years to as much as four years. DOE attributes the problem to several factors, including rising electricity demand, aging infrastructure, workforce challenges and shortages of essential raw materials.⁴

The magnitude of the demand increase is striking. Wood Mackenzie reported that US demand for generator step-up transformers increased 274% between 2019 and 2025. Reuters independently reported the same Wood Mackenzie data in May while documenting developers scrambling for factory slots and imports as supply struggled to keep up.⁶

Manufacturers are adding capacity, so there’s no reason to assume today’s shortage lasts forever. For now, though, an advanced reactor project can find itself waiting years for electrical equipment that has nothing particularly nuclear about it.

You can improve the reactor all you want. It doesn’t make the transformer arrive any faster.

3. Grid access

The third constraint becomes particularly clear when you look at Three Mile Island.

Most people have heard of Three Mile Island for one reason, and it isn’t pretty. In 1979 it was the site of the most serious accident in US commercial nuclear power history, when a combination of equipment failures, design problems and operator errors led to a partial meltdown of one of its reactors.⁷

So, understandably, the idea of restarting it may sound like the worst idea ever. There is an important distinction, though. The accident happened at Unit 2. The reactor Constellation is now trying to restart is Unit 1, a separate reactor that continued operating for another 40 years before shutting down in 2019 for economic reasons.⁷

A lot changed after the accident too. Three Mile Island led to major changes in US nuclear regulation and operations, including reactor design requirements, operator training, emergency preparedness and NRC oversight. Before Unit 1 can restart, the NRC still has to approve the necessary licensing changes, inspect restored and new equipment and determine that the plant is ready to return to operation.⁷

Constellation expects to have Unit 1, now called the Crane Clean Energy Center, ready to produce electricity in 2027 under a 20-year power purchase agreement with Microsoft.⁸

The grid may not be ready for it.

PJM’s initial system impact study identified transmission upgrades required for the plant to be fully deliverable to the grid, with some projected completion dates extending as late as December 2030. Constellation is pursuing alternatives that could reduce the upgrades required, so 2030 isn’t necessarily where this ends.⁹ Still, an existing nuclear power plant could be ready to produce electricity years before some of the transmission infrastructure needed to fully deliver its power.

This isn’t just a Three Mile Island problem. At the end of 2025, roughly 8,200 proposed projects representing 1,312 gigawatts of generation and another 749 gigawatts of storage were actively seeking US grid interconnection. Most projects that enter these queues are ultimately withdrawn, so those numbers shouldn’t be confused with future generating capacity. They do give you a sense of how much proposed capacity is trying to find its way onto the grid.¹⁰

For a new advanced reactor, getting the nuclear technology right is only one part of getting electricity to a customer.

So where does this leave advanced nuclear?

The takeaway isn’t that building advanced reactors has suddenly become easy. It hasn’t. Reaching criticality is an important milestone, but there’s still a long way between demonstrating that a reactor works and deploying it commercially at scale.

What caught my attention is how bottlenecks are now migrating to some specific constraints beyond the reactor itself. For better or worse, that’s where we are right now, in this moment in history.

If advanced nuclear starts scaling, the United States will need more HALEU and more capacity to deconvert it. Reactor developers will be competing for major electrical equipment that can already take years to procure. Then those reactors will need connections to a transmission system where even an existing nuclear plant can run into years of additional infrastructure work.

There are reasons for optimism. DOE is funding new HALEU enrichment and deconversion capacity. Transformer manufacturers are expanding production. Grid operators are changing interconnection processes.² ³ ⁶ ¹⁰ None of that guarantees these constraints will be resolved quickly enough, or that new ones won’t emerge as deployment grows. Bottlenecks have a way of moving as you solve them.

Still, it’s progress. A few years from now we may look back and find that some of the advances that mattered most weren’t breakthroughs in nuclear physics at all. They may have been new fuel capacity, more transformers, or a faster path onto the grid.

For now, those are some of the things that have to go right.

Endnotes

1. U.S. Department of Energy, Office of Nuclear Energy, “Fact Sheet: The Golden Era of American Nuclear Energy Has Arrived,” 2026. DOE reports that Antares Nuclear’s Mark-0 reached criticality June 4, Valar Atomics’ Ward 250 on June 18, Deployable Energy’s Unity on June 30 and Aalo Atomics’ Aalo-X on July 4. DOE describes the demonstrations as zero-power fueled criticality demonstrations and confirms the goal of at least three reactor criticalities by July 4, 2026.

2. U.S. Department of Energy, Office of Nuclear Energy, “HALEU Enrichment Services.” DOE defines HALEU as uranium enriched between 5% and 19.75%, describes US commercial HALEU enrichment services as limited and documents the January 5, 2026 task orders of $900 million each to American Centrifuge Operating, a Centrus subsidiary, and General Matter to expand domestic HALEU enrichment capacity.

3. U.S. Government Accountability Office, Nuclear Fuel: Actions Needed to Enhance Cost Reporting and Economic Analysis for Federal Uranium Supply Efforts, GAO-26-107385, 2026. GAO reports that, as of April 2026, DOE projected 21.2 metric tons of uranium in HALEU available for advanced reactor demonstration companies by the end of 2028 and 23.4 metric tons by the end of 2030. DOE officials told GAO that demand would likely exceed what DOE could supply before commercial HALEU becomes available. GAO also describes deconversion as the process following enrichment, reports that DOE contracted with six companies eligible to compete for HALEU deconversion work and states that the capability does not yet exist for HALEU. GAO separately reports that fuel fabrication was generally not viewed as a limitation by stakeholders.

4. U.S. Department of Energy, Office of Electricity, “Distribution Transformer Convening Webinar,” March 2026. DOE reported that lead times for large transformers serving substations and generators had grown from approximately three years to as much as four years and cited rising electricity demand, aging infrastructure, workforce challenges and shortages of essential raw materials among the contributing factors. DOE’s 2024 Large Power Transformer Resilience Report to Congress independently documented manufacturing lead times reaching and exceeding 36 months and described large power transformers as potential bottlenecks to rapid grid expansion.

5. Lawrence Berkeley National Laboratory, Queued Up: 2026 Edition, Characteristics of Power Plants Seeking Transmission Interconnection as of the End of 2025. Berkeley Lab reports approximately 8,200 projects actively seeking US grid interconnection at year-end 2025, representing 1,312 GW of generation and approximately 749 GW of storage. The report cautions that most projects entering interconnection queues are ultimately withdrawn.

6. Wood Mackenzie, “Mind the Gap: Tackling Supply-Chain Challenges in the Electric T&D Sector,” 2025; Laila Kearney, Reuters, “US Power Transformer Buyers Scramble for Imports, Factory Slots,” May 11, 2026. Wood Mackenzie reports that demand for transmission and distribution equipment grew between 35% and 274% from 2019 through 2025 depending on equipment type. Reuters, citing Wood Mackenzie, reports a 274% increase in demand for generator step-up transformers and documents extended lead times and manufacturers investing billions in additional capacity.

7. Constellation Energy, second-quarter 2025 results and subsequent company disclosures. Constellation states that the Crane Clean Energy Center, formerly Three Mile Island Unit 1, is expected to return to service in 2027. The restart is supported by a 20-year power purchase agreement with Microsoft and remains subject to regulatory approvals.

8. Constellation Energy Corporation, Form 10-Q for the quarter ended March 31, 2026, filed with the U.S. Securities and Exchange Commission. Constellation states that PJM’s Phase I System Impact Study identified contingent transmission upgrades required for Crane to be fully deliverable to the grid, with some projected in-service dates extending as late as December 2030. Constellation also disclosed that it sought a FERC waiver to transfer capacity interconnection rights from Eddystone to Crane in an effort to reduce the number of contingent upgrades required before full deliverability.