TerraPower’s Natrium reactor gets NRC approval: why Bill Gates’ next-gen nuclear bet just cleared its biggest hurdle

AI generated image for TerraPower’s Natrium reactor gets NRC approval: why Bill Gates’ next-gen nuclear bet just cleared its biggest hurdle

On March 5, 2026, The Verge ran the kind of headline that makes both climate hawks and electricity planners sit up straighter: Bill Gates’ nuclear company, TerraPower, is the first to get federal approval to build a next-generation reactor in the United States. The piece—written by Jess Weatherbed—tracks a pivotal milestone: the U.S. Nuclear Regulatory Commission (NRC) has approved a construction permit for TerraPower’s first commercial-scale Natrium plant near Kemmerer, Wyoming, with completion targeted around 2030.

If you’ve followed advanced nuclear for the last decade, you’ll know why this matters: plenty of companies have glossy renderings and confident timelines; far fewer have a green light from the regulator that famously treats “move fast and break things” as a personal insult. This approval doesn’t mean the plant is finished or even operating—it means the project can build the nuclear portions under the NRC’s construction permit framework. But in nuclear, that’s like clearing the “boss level” before you even pick up the legendary sword.

Below is what was approved, why Natrium is different, what still has to happen before electrons hit the grid, and why the broader energy and tech industries—yes, including AI data centers—are paying attention.

What exactly got approved—and when?

The key event happened on March 4, 2026: the NRC approved a construction permit for TerraPower’s Kemmerer Power Station Unit 1 project, which uses the company’s Natrium sodium-cooled fast reactor technology. Multiple outlets reported the decision as the NRC’s first construction permit for a commercial reactor in roughly eight years—and notably, the first for a non-light-water commercial reactor in the U.S. in more than 40 years. The Associated Press characterized it as a significant regulatory milestone for a Bill Gates-backed project in Wyoming.

The Verge article published the next morning, March 5, 2026, and frames TerraPower as the first company to receive federal approval to build a commercial-scale advanced reactor plant—while tying the moment to strained grids and rising demand from AI infrastructure. (We’ll get to that “AI meets nuclear” subplot shortly.)

It’s worth being precise about what “approval” means here. The NRC has been reviewing TerraPower’s construction permit application since it was submitted in 2024, with formal docketing and safety/environmental review steps along the way. The construction permit is a permission slip to build the facility as designed, subject to regulatory requirements and inspections. It is not an operating license. A separate approval process is still required before the plant can load fuel and run.

Quick primer: what is the Natrium reactor?

Natrium is a sodium-cooled fast reactor design, developed by TerraPower with GE Hitachi as a key technology partner (TerraPower and DOE describe this partnership in background materials and program documents). Instead of using water under high pressure as the primary coolant, Natrium uses liquid sodium to carry heat away from the reactor core. The Natrium demonstration plant is widely described as a 345 megawatt electric (MWe) reactor that can deliver up to 500 MWe during peak demand periods by using integrated thermal energy storage.

The Verge summary captures the design’s most headline-friendly elements: sodium cooling, lower-pressure operation, and a molten salt-based energy storage system that enables flexible output. That last bit matters because it’s one of the most practical “grid-era” twists in the advanced reactor world: it’s not only about making nuclear smaller or different; it’s about making it behave more like a dispatchable resource in a grid full of variable wind and solar.

Sodium cooling: the “not water” choice (and why it’s complicated)

Most commercial reactors today are light-water reactors (LWRs), meaning ordinary water does double duty: it cools the core and helps control the nuclear reaction. Water is familiar, heavily engineered, and well understood by regulators and operators.

Sodium-cooled fast reactors take a different route. Sodium is an efficient heat transfer medium and allows operation at lower pressure than water-cooled designs—one of the reasons supporters argue it can reduce certain engineering burdens. But sodium brings its own personality: it reacts vigorously with water and burns in air, which means the plant design must take sodium handling, leak detection, and fire suppression seriously. The technology isn’t “new” in a global sense—there have been sodium-cooled reactors in various countries and in experimental settings—but it is new to the U.S. commercial licensing pipeline at this scale.

The molten salt energy storage: nuclear that can “sprint”

Natrium’s standout feature is its coupling of a nuclear heat source with a molten salt-based thermal storage system. The basic idea: the reactor can run steadily, while heat can be stored and later used to generate extra electricity when the grid needs it most.

The Associated Press report on the March 4, 2026 approval describes the plant as 345 MWe with the ability to produce up to 500 MWe at peak. The Verge likewise notes the “boost” capability. That’s not the same as a battery, and it’s not magic; it’s a thermal storage system designed to help with load following and peak support without constantly ramping the nuclear core up and down.

In practical grid terms, this aims to answer a real question utilities face as renewables grow: how do you keep reliability when the sun is not cooperating and the wind is taking a nap?

Why Kemmerer, Wyoming?

The Natrium project is being developed near Kemmerer, Wyoming, adjacent to existing fossil generation infrastructure in the region. TerraPower and its utility partner, PacifiCorp, have long positioned the project as part of a broader “coal-to-nuclear” transition strategy: keep an energy community producing energy, keep skilled workers in the region, and reuse grid connections and industrial know-how rather than starting from scratch in a greenfield site.

This theme shows up repeatedly in public reporting and project descriptions. The Associated Press has described the Natrium plant as set to be built near a coal-fired plant that is being converted to burn natural gas outside Kemmerer. That context is important because it underlines why advanced nuclear developers often chase retiring coal plant sites: transmission is already there, the land is already industrial, and the community’s identity (and tax base) is already tied to power generation.

The “energy transition” realism: reuse what you can

In a perfect world, we’d decarbonize with minimal controversy, unlimited skilled labor, and power lines that build themselves. In the real world, siting and transmission are hard. Retiring coal sites are one of the few places where you can plausibly build large clean firm power without starting every local conversation from zero.

That doesn’t mean it’s frictionless. Nuclear projects still face scrutiny, local concerns, workforce questions, and long timelines. But compared with brand-new greenfield projects, coal-to-nuclear conversions can look like the “path of least impossible.”

How big is this milestone for the NRC—and for advanced nuclear?

Regulatory progress is often the quiet bottleneck in advanced nuclear. Many designs have been stuck in the “we’re totally ready… once someone funds, licenses, and builds it” stage for years.

That’s why the March 4, 2026 NRC approval matters beyond TerraPower. According to the Associated Press, this is the NRC’s first construction permit for a commercial reactor in eight years, and its first approval for a non-light-water commercial reactor in more than 40 years. That’s not just a TerraPower story; that’s a signal that the U.S. regulator can, in fact, process and approve a commercial advanced reactor construction permit in the modern era.

It also suggests that the NRC’s processes for advanced reactors—often criticized as being too slow or too anchored to LWR assumptions—can produce a tangible outcome. The debate will continue (it always does), but the “nobody can get anything approved” narrative just took a hit.

What still has to happen before the reactor can operate?

Even with a construction permit, TerraPower still must clear additional regulatory steps before operation. Construction permits are not operating licenses. The plant must be built to specification, inspected, tested, and then separately authorized to load fuel and begin operations under NRC oversight.

In short: the project is now allowed to build the nuclear plant, but it is not yet allowed to turn it on.

Construction risk is still real (ask Plant Vogtle)

If you want a cautionary tale that lives rent-free in every U.S. nuclear discussion, it’s Plant Vogtle in Georgia. The expansion completed the first new, scratch-built U.S. reactors in a generation, but it also came with major cost overruns and schedule delays. The Associated Press referenced Vogtle’s roughly $35 billion total price tag and significant overruns in earlier reporting about TerraPower’s project timeline and ambitions.

Natrium is different in design and scale, and TerraPower argues future units should be cheaper after first-of-a-kind engineering and licensing are done. Still, a first deployment is, by definition, a first deployment. Anyone pretending construction risk has been “solved” is selling something—likely a PDF.

The AI data center angle: why Big Tech is watching

The Verge frames this project in the context of electrical grids “under strain from AI data centers.” That’s not just a convenient buzzword mashup. Data centers are expanding rapidly, and AI workloads can be power-hungry. Utilities and grid operators are increasingly focused on how to deliver large amounts of reliable power while meeting decarbonization goals.

Advanced nuclear developers see an opening: offer clean firm power that can run around the clock, plus (in Natrium’s case) additional flexibility via thermal storage. Meanwhile, some tech leaders have publicly discussed nuclear as a tool for climate and reliability, and Bill Gates has repeatedly advocated for advanced nuclear development through TerraPower and his public commentary.

But here’s the rub: hyperscalers and data center developers don’t just need “clean.” They need predictable timelines, bankable costs, and credible delivery. The NRC permit approval helps with credibility, but it doesn’t automatically make project finance easy or construction fast.

Fuel: the HALEU problem that delayed everything

Advanced reactors often need a different fuel supply chain than existing LWR fleets. Many designs—including Natrium—are associated with HALEU (high-assay low-enriched uranium), generally enriched to between 5% and 20% U-235. The problem: for years, commercial-scale HALEU availability has been constrained, with Russia playing an outsized role in supply.

TerraPower publicly delayed its Natrium timeline after Russia’s invasion of Ukraine disrupted assumptions about fuel sourcing. The Union of Concerned Scientists published a statement noting TerraPower pushed the expected operation timeline back and highlighted the lack of feasible alternative HALEU suppliers at the time. Broader reporting has also noted the strategic vulnerability of relying on Russian-origin nuclear fuel services, particularly for next-gen designs.

Since then, the U.S. has taken steps to reduce reliance on Russian uranium imports, and policymakers have increasingly treated domestic enrichment and HALEU availability as a national strategic issue. But building fuel supply chains takes time, and advanced reactor deployment schedules will continue to be linked to fuel availability, licensing of fabrication, and the ramp-up of enrichment capacity.

Why this matters for “hundreds of reactors” ambitions

It’s easy to say “we plan to build hundreds of reactors.” It’s much harder to source fuel for hundreds of reactors—especially if that fuel is not the same as what the existing fleet uses, and especially if your preferred supply chain is geopolitically complicated.

So while the NRC construction permit is a major de-risking event on the regulatory side, the supply chain side (fuel, components, qualified labor, and specialized materials) remains a parallel challenge.

The policy and funding context: ARDP, cost share, and why government is involved

TerraPower’s Natrium project is one of the Department of Energy’s flagship advanced reactor demonstrations supported under the Advanced Reactor Demonstration Program (ARDP). DOE materials describe ARDP as cost-shared partnerships intended to help get advanced reactors demonstrated on an aggressive timeline, with TerraPower (Natrium) and X-energy (Xe-100) as major early selections.

DOE announced initial awards in 2020, positioning the program as an effort to demonstrate advanced reactors that could be operational within about seven years. Over time, both projects have faced the real-world challenges of licensing, construction complexity, and supply chain constraints. But the underlying logic—public-private cost sharing to move first-of-a-kind energy infrastructure from PowerPoint to concrete—has remained consistent.

Why cost share is not a dirty word (in infrastructure)

In software, a beta version can be deployed, patched, and iterated quickly. In nuclear, “ship now, fix later” is not a strategy; it’s a scandal waiting to happen. First-of-a-kind demonstrations carry high upfront costs precisely because you are establishing new licensing precedents, qualifying new components, and building the industrial ecosystem around the design.

That’s one reason ARDP-style cost sharing exists: it socializes some of the innovation risk for technologies that—if they work—could deliver broad public benefits such as decarbonization, reliability, and energy security. It’s not unique to nuclear; we do similar things for other strategic technologies. The difference is that nuclear projects are big, visible, and allergic to delays (even though they still happen).

How Natrium compares to other “next-gen” nuclear approaches

Advanced nuclear is a broad umbrella, and it’s worth mapping Natrium’s niche relative to its peers:

  • Large LWRs (traditional): Proven technology and existing supply chain, but recent U.S. builds have struggled with cost and schedule risk.
  • SMRs (light-water): Aim for smaller units and more factory-style construction. NuScale’s design certification milestone (for example) showed progress on the regulatory front for SMRs, even as market and project execution challenges remain for specific deployments.
  • High-temperature gas reactors (like X-energy’s Xe-100): Often paired with industrial heat use cases and TRISO fuel development; also supported under ARDP in parallel with Natrium.
  • Microreactors: Very small reactors aimed at remote sites, defense applications, or niche industrial uses—different market than a 345 MWe grid plant.
  • Fast reactors (like Natrium): Higher-temperature, different coolant, and different fuel considerations; historically promising, but commercially elusive in the U.S. until now.

Natrium’s “killer feature,” if it has one, is the blend of nuclear with thermal storage to provide grid flexibility. That’s a direct response to modern grid needs rather than a purely academic design choice.

Why the NRC’s decision could ripple through the industry

For the advanced nuclear sector, regulatory approval is a form of validation that affects more than one project:

  • Investors: A credible regulatory pathway reduces perceived risk (though it does not remove construction risk).
  • Utilities: A licensed, permitted project is easier to consider in resource planning than a design that is still “in pre-application discussions.”
  • Supply chain: Vendors are more likely to invest in capability when there’s a real project with a real schedule.
  • Regulators and policymakers: The decision becomes a reference point for how future advanced reactor applications might proceed.

GeekWire’s reporting on the approval includes comments from TerraPower leadership that emphasize ambitions for large-scale deployment beyond the first unit, reinforcing the notion that Kemmerer is intended as a market opener, not a one-off science fair.

Concerns and criticisms: safety, waste, cost, and public trust

No nuclear story is complete without the four horsemen of the comment section: safety, waste, cost, and “why not just build more renewables.” These are not frivolous questions. They are the reason nuclear regulation exists and the reason projects take a long time.

Safety: advanced designs still have to prove operational excellence

Natrium’s sodium coolant changes the engineering trade space. Supporters argue the lower-pressure coolant loop and thermal characteristics can provide safety advantages. Critics point out sodium’s chemical reactivity introduces other hazards that must be engineered and managed. Either way, safety is not a marketing claim; it’s a performance record built over years of operation, inspections, and transparency.

Waste: not solved by vibes

Advanced reactors do not eliminate nuclear waste. They may change the waste profile, and some fuel cycle concepts aim to improve long-term management. But the U.S. still lacks a permanent geological repository for commercial spent fuel, and interim storage remains the reality. Any serious buildout of nuclear—advanced or conventional—has to grapple with that policy gap.

Cost and schedule: the make-or-break test

Even nuclear supporters will admit the industry’s modern challenge is not physics; it’s project execution. The Natrium demonstration will be watched closely as a test of whether advanced nuclear can avoid the worst cost and schedule pitfalls that have haunted large projects. The AP report pegs completion around 2030, and TerraPower has publicly communicated similar targets.

If Natrium comes in close to schedule and budget, it becomes a blueprint. If it slips badly, it becomes another cautionary tale—this time with molten salt storage and sodium coolant as the plot twist.

What happens next: a realistic timeline

Based on public reporting and project statements, TerraPower’s target is to complete construction around 2030. The March 4, 2026 construction permit makes it plausible for major nuclear construction to proceed, but “plausible” is not “guaranteed.”

A realistic set of next steps looks like this:

  • Detailed engineering and procurement: Lock in long-lead components and qualified vendors.
  • Site work and non-nuclear construction: Already underway in parts since 2024, according to reporting and project updates.
  • Nuclear island construction under NRC oversight: The main “now we’re really doing it” phase enabled by the construction permit.
  • Fuel supply chain readiness: Secure HALEU supply and any necessary fuel fabrication approvals.
  • Testing, commissioning, and operating license steps: Separate from the construction permit and essential for operation.

In other words, the hardest regulatory hurdle to begin building the nuclear plant has been cleared, but the project still has several years of complex work ahead—and a few remaining dragons to fight, including fuel.

The bigger picture: does this make advanced nuclear “real” again in the U.S.?

“Real” is a high bar in energy infrastructure. But the NRC decision is undeniably a step toward reality.

For decades, the U.S. nuclear story has been stuck between nostalgia (“we built big reactors once”) and aspiration (“this next design will fix everything”). What’s different now is a convergence of pressures that make clean firm power attractive again:

  • Decarbonization targets that require more than just variable renewables
  • Grid reliability needs in the face of extreme weather and electrification
  • Load growth driven by electrification, reshoring, and data centers
  • Energy security concerns that make fuel supply chains strategic

TerraPower’s Natrium project sits at the intersection of those forces. It’s also a reminder that, in energy, “the future” is not announced—it’s permitted, financed, constructed, fueled, and operated. Preferably in that order.

Sources

Bas Dorland, Technology Journalist & Founder of dorland.org