Why Nuclear Is Back

Nuclear energy never disappeared from the American grid. The United States still operates the world’s largest commercial nuclear fleet by capacity, with 94 operating reactors and roughly 97 gigawatts of net capacity. Those plants routinely produce around one-fifth of U.S. electricity.

What disappeared was confidence in building new reactors. Vogtle Units 3 and 4 were the first new U.S. reactors brought into commercial operation in decades. Their completion proved that new large light-water reactors could still be built in America.

The cost proved why few utilities wanted to try again. Today, the demand environment is different.

For much of the previous decade, U.S. electricity demand was relatively flat.

Now utilities are planning for AI data centers, new semiconductor fabs, advanced manufacturing, electrified transportation and population growth. At the same time, existing coal plants are aging, gas pipelines face regional constraints, renewable generation is weather-dependent and transmission expansion remains slow.

Nuclear re-enters the conversation because it provides high-capacity-factor generation without requiring continuous fuel delivery or large land areas. The key word is firm.

The grid does not only need megawatt-hours over a year. It needs megawatts at specific hours. That is why data-center developers are increasingly willing to pay for nuclear attributes or support projects directly.

SMR, Advanced Reactor and Microreactor Are Not Synonyms

Nuclear terminology has become marketing shorthand. It needs discipline.

A small modular reactor describes scale and modularity. An advanced reactor generally describes technology that differs materially from conventional operating light-water reactors through coolant, fuel, architecture or safety approach.

A microreactor describes an even smaller system, often measured in single-digit or tens of megawatts and aimed at remote, military, industrial or specialized uses. These categories overlap.

A microreactor may also be an advanced reactor. An SMR may use conventional light-water technology. An advanced reactor may not be especially small.

Any claim that “SMRs are safer” or “advanced reactors are cheaper” should therefore trigger a follow-up question: Which design? This point matters because the technology should be evaluated as part of the surrounding system rather than as an isolated claim or capability.

TerraPower’s Natrium Crossing

TerraPower’s Natrium project is one of the most important advanced-reactor developments in the United States because it has crossed a regulatory threshold that many concepts never reach. On March 9, 2026, the NRC issued a construction permit for Kemmerer Power Station Unit 1 in Wyoming.

The NRC describes Natrium as a 345-megawatt-electric pool-type sodium fast reactor using HALEU metallic fuel. The design also incorporates a separate molten-salt energy-storage system.

That architecture matters because it decouples reactor thermal output from grid electrical output. The reactor can operate steadily while heat is stored.

When grid demand rises, stored heat can supplement reactor heat to increase turbine output. This is an unusual bridge between Article 012’s energy-storage problem and nuclear generation.

Natrium is not yet an operating commercial reactor. A construction permit is not an operating license.

The project must still complete construction, fuel qualification, commissioning, operating-license requirements and commercial startup. But the permit matters because it demonstrates that the U.S. regulatory system can authorize construction of a commercial non-light-water reactor.

Canada Builds the First BWRX-300

The most concrete commercial SMR construction story in North America is currently in Ontario. Ontario Power Generation selected GE Vernova Hitachi’s BWRX-300 for the Darlington New Nuclear Project.

The Canadian Nuclear Safety Commission issued a construction license in April 2025. In March 2026, the regulator removed the first mandatory hold point after determining OPG had satisfied commitments necessary for reactor-building foundation work.

The CNSC describes the project as one BWRX-300 under construction, with the possibility of additional units later. The reactor is a roughly 300-megawatt boiling-water design intended to simplify systems compared with previous generations.

The economic hypothesis is straightforward: use a smaller reactor; standardize the design; build several; keep the construction organization together; and learn from repetition. That is the theory of economies of multiples.

The important evidence will come from the construction sequence. If Unit 2 is faster than Unit 1, and Unit 3 faster still, the SMR model gains credibility. If every unit behaves like another bespoke megaproject, smaller size alone will not save the economics.

X-energy, TRISO and Industrial Heat

X-energy’s Xe-100 attacks a different market. The reactor is a high-temperature gas-cooled pebble-bed design using TRISO fuel.

Each Xe-100 module is designed around roughly 80 megawatts of electricity and about 200 megawatts of thermal output. The thermal number matters.

Many industrial facilities need steam or high-temperature process heat rather than electricity alone. Dow’s proposed Seadrift, Texas project would use four Xe-100 modules at an industrial manufacturing site.

The NRC accepted the project’s construction-permit application in 2025. The project is significant because nuclear would be integrated directly into industrial production rather than operating only as a remote grid power plant.

Fuel is central to the architecture. TRISO fuel encloses very small uranium particles within multiple ceramic and carbon layers designed to retain fission products at high temperatures.

X-energy’s TRISO-X subsidiary received an NRC special nuclear material license for its commercial fuel-fabrication facility in Tennessee in 2026. DOE says fuel fabrication is expected to begin in 2028. That timeline illustrates a recurring lesson: an advanced reactor is not deployable until its fuel factory is deployable.

Kairos, Google and the Road From Test Reactor to Power Plant

Kairos Power’s path is intentionally incremental. Rather than jumping immediately from design work to a large commercial plant, the company has pursued a sequence of test reactors and demonstrations centered on fluoride-salt-cooled high-temperature technology.

The NRC lists Kairos’ Hermes and Hermes 2 projects among advanced research and test reactors under licensing activity. Google’s partnership with Kairos brings a commercial customer into that development path.

In 2024 Google announced an agreement for up to 500 megawatts from multiple Kairos reactors. In 2025 Google, Kairos and the Tennessee Valley Authority announced a more concrete first deployment: TVA would purchase electricity from the Hermes 2 plant in Oak Ridge, with 50 megawatts intended to support the grid serving Google data-center demand in Tennessee and Alabama beginning around 2030.

This is more than a power purchase agreement. It creates a three-party commercialization structure:

That structure may become important across advanced energy. Developers need customers.

Utilities need confidence.

Customers want power but do not necessarily want to own nuclear plants. Long-term corporate demand can bridge those interests.

Microreactors Become Real Hardware

The most surprising advanced-nuclear development of 2026 may be occurring at the smallest end of the market. DOE’s Reactor Pilot Program moved several privately developed test reactors to zero-power criticality on compressed schedules.

Antares Nuclear’s Mark-0 achieved criticality in June. Valar Atomics’ Ward 250 followed later that month.

Deployable Energy’s Unity reached criticality at Idaho National Laboratory at the end of June. Aalo Atomics’ Aalo-X reached DOE’s milestone in early July.

DOE described the sequence as four advanced reactors reaching zero-power criticality by the July deadline. These milestones require precise interpretation.

Criticality means a reactor achieved a self-sustaining nuclear chain reaction. Zero-power criticality means the test is performed at very low thermal power, primarily to validate reactor physics and behavior.

It does not mean:

commercial electricity;

full-temperature operation;

multi-year reliability;

mass manufacturing;

or bankable economics.

But it is still a major step beyond simulation. The core contains fuel.

The chain reaction occurs.

Measurements can be compared against models. The industry begins generating real reactor data.

The HALEU Bottleneck

Advanced reactor discussions often focus on reactor geometry. The more immediate constraint may be fuel.

Most conventional U.S. commercial reactors use low-enriched uranium below 5 percent uranium-235.

Many advanced reactor designs require high-assay low-enriched uranium—HALEU—enriched between 5 and 19.75 percent. The additional enrichment allows smaller cores, different fuel cycles and higher burnup in some designs.

But it creates an industrial dependency. DOE says domestic commercial HALEU enrichment services remain limited.

In January 2026, DOE awarded $2.7 billion in enrichment task orders over the coming decade, including $900 million each to American Centrifuge Operating and General Matter for HALEU enrichment and $900 million to Orano Federal Services for expanded LEU enrichment. In May, NNSA announced the transfer of 1.7 metric tons of HALEU from Japan to the United States—the largest international uranium shipment in NNSA history—intended to support the emerging advanced-reactor industry.

DOE is also developing transport packages, deconversion capacity, fuel fabrication and allocation systems. This is the real nuclear fuel stack:

The Regulatory System Changes

Nuclear development has always been shaped as much by licensing as engineering. In March 2026, the NRC issued Part 53, a new optional licensing framework designed specifically to accommodate advanced reactor technologies through a more technology-inclusive, risk-informed approach.

The rule matters because the existing regulatory structure was built primarily around large light-water reactors. A sodium reactor, molten-salt system or microreactor may have different hazards, inventories, passive systems and source terms.

A modern regulatory framework needs to preserve safety requirements without forcing every design into assumptions created for a different technology. DOE has also created faster experimental pathways under its own legal authority for reactors located within DOE jurisdiction.

The Reactor Pilot Program, DOME test bed and Nuclear Energy Launch Pad are examples. The distinction is important:

DOE authorization of a test reactor is not the same legal pathway as NRC licensing of a commercial power reactor. Successful testing may produce evidence useful for later commercial applications. It does not eliminate the commercial licensing requirement.

Why Big Tech Wants Nuclear

The arrival of Microsoft, Google, Meta and Amazon in nuclear power changes the financing landscape. Technology companies are unusual electricity customers.

They can sign long contracts.

They have high credit quality.

They can tolerate some first-of-a-kind technology risk. They need enormous amounts of firm power.

And their future electricity demand may grow faster than traditional utility planning cycles. The approaches differ.

These arrangements should not be described as proof that advanced nuclear is economically solved. They are evidence that a new class of customers is willing to help solve it.

The Cheapest New Nuclear Plant May Be an Old One

The nuclear renaissance is not only about new designs. In the near term, the fastest nuclear megawatt may come from preserving or restarting existing plants.

Microsoft’s agreement with Constellation supports the restart of the former Three Mile Island Unit 1, now renamed the Crane Clean Energy Center. Microsoft says the 835-megawatt facility is expected to supply steady nuclear power associated with its data-center demand beginning in 2027.

DOE has also financed the restart of Palisades in Michigan. Meta’s Clinton agreement performs a different function: it helps keep an existing reactor operating for another two decades and supports an uprate that adds output.

The economics are intuitive.

An existing nuclear site already possesses: grid interconnection; transmission; cooling infrastructure; security; a licensed site; trained workforce; roads; community experience; and much of the physical plant. This is why restarts and life extensions deserve to be considered part of the new nuclear race.

They are not technologically glamorous. They may deliver power sooner than many advanced reactors.

The Economic Reality Check

Nuclear advocates have good reasons to be excited about today’s activity. They also have recent history to overcome.

Vogtle Units 3 and 4 illustrate both sides. The two AP1000 reactors now provide more than two gigawatts of new nuclear capacity and are expected to operate for decades.

EIA reports that construction began in 2009, with the units originally expected to cost around $14 billion and enter operation in 2016 and 2017. Instead, commercial operation came in 2023 and 2024 and total project cost exceeded $30 billion.

The lesson is not that nuclear power cannot work. The reactors are working.

The lesson is that project execution can dominate reactor physics. NuScale’s cancelled Carbon Free Power Project provides the SMR version of the warning.

NuScale had achieved major regulatory progress. But UAMPS and NuScale terminated the project in 2023 after it became unlikely that enough customers would subscribe to continue.

NuScale had earlier raised the project’s target power price to $89 per megawatt-hour amid higher commodity and construction costs. A 2026 DOE Inspector General audit examined lessons from federal oversight of the terminated demonstration project. This is why small size alone does not guarantee cheap nuclear power.

The Global Nuclear Technology Race

The United States is not developing advanced nuclear in isolation. The International Atomic Energy Agency’s PRIS database shows more than seventy power reactors under construction globally.

China remains especially important because it is simultaneously building large conventional reactors and an SMR. The Linglong One / ACP100 at Hainan Changjiang is a 100-megawatt-electric pressurized-water SMR under construction. Current 2026 reporting indicates it is moving through pre-commissioning work.

China also connected new gigawatt-scale reactors to the grid in 2026 and continues construction across multiple sites. Canada’s Darlington project may become one of the most consequential Western SMR demonstrations because it uses a design other countries are considering and because Ontario intends the site to support multiple units.

The OECD Nuclear Energy Agency says it is now tracking roughly 129 SMR designs globally, though many are at very different levels of maturity and some are paused or inactive. That number should not be interpreted as 129 future commercial products.

It demonstrates how crowded the design race has become. The eventual market may consolidate sharply around designs that can actually clear licensing, secure fuel, finance construction and build fleets.

Microreactors and National Security

The military value proposition for microreactors is different from the utility value proposition for gigawatt nuclear plants. Military installations want resilience.

Remote installations may depend on vulnerable fuel deliveries or long transmission lines. Critical missions require electricity even if the civilian grid fails.

DOE says it is supporting Department of War efforts including Project Pele, the Janus program, an Eielson Air Force Base pilot and Advanced Nuclear Power for Installations. Project Pele is intended as a transportable microreactor prototype for resilient military power.

The Janus effort is examining multiple installations for microreactor demonstrations. Eielson is focused on fixed-site nuclear power for remote operations.

This is a natural extension of Articles 008 and 009. Resilient communications and machine-speed military systems require resilient electricity.

A base whose sensors, satellites, cyber systems and command networks depend entirely on one external transmission corridor has an energy vulnerability. Microreactors are one possible answer. They still must prove economics, security, transportability, maintenance and operational practicality.

The Nuclear Industrial Stack

A reactor does not exist as a single technology. It sits on top of an industrial stack as deep as the semiconductor industry.

This is why nuclear strategy intersects so many earlier articles in the Critical Technology Stack. Article 006: power demand.

Article 007: materials, industrial sovereignty and advanced manufacturing. Article 011: transformers and grid interconnection.

Article 012: storage and firm-power portfolios. Nuclear power is not separate from those systems. It is one of the largest physical machines inside them.

The SURVXCOM Nuclear Deployment Test

When a nuclear developer announces a new reactor, the useful question is not whether the concept sounds advanced. Ask how much of the deployment stack actually exists.

What to Watch Next

The Return of the Reactor

Nuclear energy is returning to the center of technology policy for a reason that has little to do with nostalgia. The economy wants more electricity.

Not eventually.

Now.

AI data centers want it.

Semiconductor fabs want it.

Advanced manufacturing wants it.

Military installations want resilient versions of it. Communications networks want it.

The transmission system needs sources that can deliver when weather-dependent generation cannot. Nuclear power is one answer to that problem.

But nuclear’s history makes one thing clear. Reactor physics is not the hardest part anymore.

The difficult part is building an industrial system around the reactor. Fuel.

Licensing.

Factories.

Heavy components.

Construction.

Financing.

Workforce.

Transmission.

Waste.

Repeat orders.

Every advanced reactor company ultimately has to solve those ordinary industrial problems. The industry’s new optimism is justified by real developments:

commercial permits;

construction;

new fuel facilities;

criticality demonstrations;

new regulatory pathways;

and customers with enormous electricity needs. But the final test is brutally simple.

Can advanced nuclear make the transition from a technology that can be built to a technology that can be built repeatedly? That is where the nuclear race will be won.

The future of nuclear power will not be decided by the most impressive reactor rendering. It will be decided by the reactor industry that learns how to build the same machine again.