Key Highlights

• Valar Atomics partnered with Nvidia to power a data center using a microreactor in Utah, in what the companies describe as a first-of-its-kind demonstration.

• The reactor is cooled with helium rather than water, addressing rising scrutiny of data center water consumption.

• Valar is among roughly ten startups in a Department of Energy pilot program targeting small reactor criticality by July 4.

A First-of-Its-Kind Demonstration Amid Rising Infrastructure Scrutiny

Valar Atomics, a California-based nuclear startup, said it powered a small data center in Utah using its microreactor technology, running a live demonstration with Nvidia's Blackwell chip architecture. The companies described it as the first instance of a small reactor directly powering an AI data center facility. The announcement arrives as public and political attention increasingly focuses on the resource demands of AI infrastructure, with a recent poll finding that only one in three Americans approve of the current pace of data center construction ahead of November midterm elections.

Valar's approach uses helium cooling rather than water, positioning the technology as a potential response to mounting concerns over the sector's water usage, which has become a politically sensitive issue in several regions facing data center buildouts.

Water and Power Constraints Reshape AI Infrastructure Design

The Valar-Nvidia demonstration coincides with Nvidia's broader push toward closed-loop liquid cooling for its next-generation DSX data center design, which the company says could reduce facility cooling water consumption from roughly 2.6 million gallons per megawatt annually to near zero. Together, these developments suggest AI infrastructure providers are increasingly treating water conservation as a core design constraint rather than a secondary consideration.

Valar is one of approximately ten nuclear startups participating in a Department of Energy pilot program with a stated goal of demonstrating three small reactors reaching criticality by July 4, reflecting federal interest in accelerating small reactor deployment timelines that have historically been measured in years rather than months. The Trump administration has separately pursued executive orders aimed at quadrupling nuclear deployment, providing a policy tailwind for companies pursuing behind-the-meter nuclear projects that bypass traditional grid interconnection and permitting processes.

Regulatory and Public Acceptance Remain Open Questions

Behind-the-meter nuclear power could offer AI infrastructure developers a path to faster deployment by avoiding lengthy grid interconnection queues, though microreactor technology remains largely unproven at commercial scale and regulatory frameworks are still evolving. Valar's involvement in litigation against the Nuclear Regulatory Commission, alongside Texas and Utah, over state versus federal licensing authority for microreactors, illustrates that regulatory uncertainty could still pose a meaningful obstacle to widespread deployment regardless of technical feasibility demonstrated in Utah.

Public sentiment data suggesting broad unease with data center expansion may complicate siting and community approval processes even for technologies marketed as water-conserving, particularly if nuclear-specific safety concerns compound existing opposition. Whether nuclear microreactors become a mainstream solution for AI power demand or remain a niche option alongside natural gas and grid power is likely to depend heavily on how quickly regulatory clarity and public trust develop alongside the technology itself.

Conclusion

The Valar Atomics and Nvidia demonstration marks an early but notable step in exploring nuclear microreactors as a potential solution to the water and power constraints facing AI data center expansion. While the technical demonstration addresses immediate resource efficiency concerns, unresolved regulatory questions and public skepticism toward data center growth suggest the path to broader adoption remains uncertain. The outcome of the Department of Energy's pilot program and ongoing licensing disputes may prove as consequential to the technology's future as the engineering demonstration itself.