Valar Atomics achieved a US first on July 1, 2026: powering an Nvidia Blackwell AI chip with a nuclear microreactor, demonstrating a breakthrough approach to addressing the escalating energy demands of AI data centers.
The Utah demonstration used Valar’s Ward 250 test reactor to power the Nvidia chip, marking the first time a US advanced reactor has powered commercial AI computing hardware. The demonstration represents a potential solution to one of the most pressing challenges facing AI infrastructure expansion—adequate and reliable power supply.
The energy problem is acute. Modern AI data centers consume enormous amounts of electricity, with Nvidia’s Blackwell GPU clusters requiring hundreds of megawatts. Traditional grid infrastructure struggles to keep pace with demand, and water requirements for cooling have become a significant constraint in many regions.
Valar and Nvidia are now jointly studying a 30-megawatt, water-free closed-loop cooling facility specifically designed to address both power and water constraints at AI data centers. The reactor produces approximately 100 kilowatts of thermal output—small by power plant standards but potentially significant for targeted AI deployment scenarios.
The company has secured $450 million in funding at a $2 billion valuation, signaling strong investor confidence in nuclear microreactors as a solution for energy-intensive computing. The demonstration validates the core technology and opens pathways for larger deployments.
The partnership addresses a critical bottleneck in AI infrastructure scaling. Hyperscalers building trillion-parameter models face real limits on where they can deploy due to power availability. Nuclear microreactors could enable AI data centers in locations previously impractical due to grid limitations.
While the Ward 250 output remains far below the scale needed for major data centers, the proof of concept demonstrates the technology’s viability. The next phase—scaling to megawatt-level outputs—could reshape how AI companies think about infrastructure placement and expansion.