Valar Atomics has reached a nuclear milestone in rural Utah that could reshape how the tech industry powers its most energy-hungry operations. The startup’s Ward 250 research reactor completed a zero-power fueled criticality demonstration at the Utah San Rafael Energy Lab in Emery County, making it the first Department of Energy–authorized reactor built outside of a national laboratory. The 250 kWth high-temperature gas-cooled reactor uses TRISO fuel, helium coolant, and a graphite moderator, and it operates under DOE authority through a categorical exclusion in the NEPA process. While no federal documents tie Nvidia or any commercial data-center customer to the project, the reactor’s design and location raise a pointed question: whether this small research unit is the opening move in a larger play to co-locate nuclear power with high-density computing infrastructure in remote locations.
A 250 kWth reactor and the race to July 4, 2026
The Ward 250 did not appear in a vacuum. It emerged from a DOE pilot program designed to expedite testing of advanced reactor designs under federal authority but outside the traditional national laboratory system. That program set an explicit target: multiple reactors achieving criticality by July 4, 2026. With Valar’s demonstration already complete, the company is ahead of that deadline by a meaningful margin.
The pilot program itself traces back to an executive order issued in May 2025, titled “Deploying Advanced Nuclear Reactor Technologies for National Security.” That order established interagency roles for the DOE and Department of Defense, framing advanced reactors not as a climate tool or a commercial energy product but as a national security priority. The distinction matters because it unlocked a faster regulatory track, one that bypasses the years-long Nuclear Regulatory Commission licensing process that has stalled other reactor projects.
Valar’s reactor sits at the Utah San Rafael Energy Research Center in Emery County, a sparsely populated area in central Utah. The DOE’s NEPA filing describes the Ward 250 as a 250 kWth high-temperature gas-cooled reactor, or HTGR. At that thermal output, the reactor cannot power a commercial data center on its own. A single AI training cluster can demand tens of megawatts of electricity. The Ward 250 produces a fraction of that. But the point of a research reactor is rarely the reactor itself. It is what comes after.
In practical terms, a 250 kWth unit is a laboratory for fuel performance, materials behavior, control systems, and safety cases. It lets engineers validate models and gather real-world data under carefully bounded conditions. If those tests support scaling up, the same design principles could be applied to reactors in the tens or hundreds of megawatts-units that begin to match the load profiles of hyperscale cloud and AI facilities.
The regulatory beachhead hypothesis
The strongest way to read the Ward 250 is as a proof of concept for a regulatory pathway, not as a power source. By building and operating the first DOE-authorized reactor outside a national lab, Valar has demonstrated that a private company can site, construct, and achieve criticality on a non-federal facility under DOE authority. That precedent did not exist before this project.
The DOE celebrated the achievement as the second advanced reactor to achieve criticality under the pilot program. The agency described the event as a “zero-power fueled criticality demonstration,” a technical term meaning the reactor sustained a controlled chain reaction without generating usable thermal output. This is a standard step in reactor commissioning, but the speed and location are what set it apart.
If the pilot program expands, the regulatory template Valar has established could allow larger reactors to be sited at remote locations chosen for their proximity to compute demand rather than existing grid infrastructure. AI data centers are increasingly being planned for areas with cheap land and low population density, precisely the kind of terrain where Emery County sits. A company that can place a reactor next to a data center, rather than relying on transmission lines from a distant power plant, gains a structural cost and reliability advantage.
The national security framing of the executive order adds another layer. Defense applications for small modular reactors include powering remote military installations and forward operating bases. The same reactor architecture that serves a military outpost could, in theory, serve a data center in a desert. The test infrastructure at the National Reactor Innovation Center in Idaho has been developed around exactly these kinds of dual-use designs, where civilian and defense needs overlap in both technology and siting.
Seeing Ward 250 as a “regulatory beachhead” helps explain why its modest power rating still matters. Once a company has demonstrated that DOE authority can cover a privately owned, non-lab reactor, it has a playbook for repeating the process with higher-output units. Each successful project makes it harder to argue that such reactors must be confined to national labs or traditional nuclear plant sites.
What federal records do not say about Nvidia or data centers
The headline connecting Valar Atomics and Nvidia to a nuclear-powered AI data center in Utah has circulated widely, but federal records tell a narrower story. The DOE’s NEPA categorical exclusion for the Ward 250 contains no reference to Nvidia, to data-center applications, or to any commercial power customer. The executive order and pilot-program texts similarly omit any discussion of AI infrastructure or Utah-specific commercial siting beyond the research reactor itself. No statements on power output scale, grid interconnection, or electricity sales agreements appear in official filings.
Two source conflicts also deserve attention. The DOE described the Ward 250 as having “successfully completed a zero-power fueled criticality demonstration,” which implies the reactor is built and operational. Yet the NEPA filing describes a project scope that includes construction, fuel loading, and testing, language that typically appears before a facility is complete. The simplest explanation is timing: NEPA documentation is often prepared and approved ahead of final build-out, while public celebration of a criticality test comes only after the hardware is in place. But the discrepancy highlights how easily readers can conflate forward-looking approvals with finished infrastructure.
Another ambiguity involves the reactor’s purpose. The NEPA document emphasizes research objectives and safety analyses, while the DOE announcement under the pilot program situates Ward 250 in a broader push for advanced nuclear technologies with national security relevance. Neither record commits to any specific end-user sector. That leaves ample room for speculation about AI and cloud computing, but speculation is not evidence. At this stage, there is no documented contract, partnership, or siting decision that would bind the reactor-or any future scaled-up derivative-to Nvidia or a particular data-center operator.
For investors, policymakers, and local communities, the distinction is crucial. A research reactor operated under DOE authority is governed by a different set of rules, oversight mechanisms, and liability structures than a commercial power reactor licensed by the Nuclear Regulatory Commission. Conflating the two could lead to misplaced expectations about timelines for delivering electricity to the grid or to private customers.
What Ward 250 does signal
Even without confirmed ties to Nvidia or any other hyperscale customer, Ward 250 sends a clear signal about the direction of advanced nuclear development. It shows that private companies can leverage federal authority to move faster than the conventional licensing process would allow, especially when projects are framed as research or national security priorities. It also shows that rural regions like Emery County can become testbeds for technologies that may eventually underpin critical digital infrastructure.
Whether that future includes reactors directly powering AI data centers in Utah or elsewhere will depend on decisions that have not yet been made: how aggressively the DOE expands its pilot program, whether Congress adjusts the balance of authority between DOE and the NRC, how industry players weigh the reputational and operational risks of nuclear co-location, and how local communities respond to the prospect of reactors next to server farms.
For now, the Ward 250 is best understood as a small but consequential step-a functioning reactor, a validated regulatory path, and a new set of possibilities for pairing advanced nuclear with the most power-hungry computing systems the world has ever built. The technology is real, the policy framework is evolving, and the commercial story is still being written.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
