What happens when the hyperscale data center buildout hits a wall that no amount of GPUs can solve? Power. In July 2026, that question has a clear answer in Valar Atomics, the three-year-old startup building small modular nuclear reactors that just proved it could run an Nvidia AI chip on its own nuclear power. The company is now in talks to raise $1 billion at a $6 billion valuation, with Sequoia expected to lead the round, according to three sources familiar with the discussions.
Valar Atomics has already closed $450 million in earlier capital at a $2 billion valuation, split between $340 million in equity and $110 million in debt. The structure of the current round is notable: part of the $1 billion has been raised at a lower valuation before the rest, reflecting a trend that is becoming increasingly common in the AI fundraising environment. Investors in the same round end up paying different prices for the same company, a reality that makes it harder for outsiders to benchmark hot startups against one another. But the headline figure remains striking: a three-fold valuation increase in roughly four months.
Why Nuclear Is Suddenly the Hottest Ticket in AI Infrastructure
The reason Valar Atomics can command a $6 billion valuation in 2026 comes down to a single arithmetic problem. Data center electricity demand is projected to grow at a pace that utility companies in most regions cannot match for years. Every hyperscaler, from Amazon Web Services to Microsoft Azure to Google Cloud, is facing the reality that new data center construction timelines are now constrained not by server availability, but by grid interconnection queues that stretch five to seven years.
Nuclear power fills that gap. Small modular reactors are factory-built, miniaturized power plants designed to be cheaper, faster to deploy, and safer than traditional gigawatt-scale nuclear plants. Valar's design uses a helium-cooled, high-temperature gas reactor, a technology path that differs from the water-cooled approaches of competitors like NuScale Power. Earlier this month, the company demonstrated its reactor powering an Nvidia AI chip in a proof-of-concept test, and simultaneously announced a partnership with Nvidia to explore nuclear energy for future AI data centers. That combination of technical validation and strategic partnership with the world's most valuable chip company gives Valar a credential that no other nuclear startup can claim.
Sequoia, Anduril, Palantir: The Unusual Investor Coalition
The investor syndicate backing Valar Atomics reads like a cross-section of American defense and technology power. Sequoia, expected to lead the new round, is the most prominent venture capital firm in Silicon Valley. But Valar's existing backers include Palmer Luckey, the founder of defense contractor Anduril, and Shyam Sankar, Palantir's chief technology officer. The defense ties are not accidental. Small modular reactors were originally developed for military applications, and the Department of Defense has been exploring SMRs for powering remote bases and data centers. Valar's legal strategy reinforces this positioning: the company joined several states and rival startups in suing the Nuclear Regulatory Commission, arguing that the agency wrongly applies the same lengthy licensing process to small test reactors that it uses for full-size commercial plants. The case has been paused repeatedly, suggesting a settlement is in the works. A more favorable regulatory framework would accelerate Valar's path to deployment by years.
The Founder Who Dropped Out at 16 and Bet on Nuclear
Isaiah Taylor, now 27, founded Valar Atomics after dropping out of high school at 16. He had launched two startups before Valar and carries a family legacy that makes his bet on nuclear power feel almost destined: his great-grandfather worked as a nuclear physicist on the Manhattan Project. Taylor's vision is audacious even by nuclear industry standards. The company plans to build hundreds of SMRs to power data centers, effectively becoming a distributed power utility for the AI industry. That level of ambition requires capital, regulatory approval, and manufacturing scale that no nuclear company has achieved in the modern era. NuScale Power, Valar's closest public competitor and the only SMR developer with U.S. regulatory design approval, has struggled with cost overruns and project cancellations. Kairos Power and TerraPower, backed by Bill Gates, are building next-generation reactors aimed at tech and industrial customers, but none have reached commercial deployment.
What the $6 Billion Bet Means for Founders
Valar Atomics is a case study in how the AI infrastructure boom is reshaping industries far beyond software. The $6 billion valuation reflects a market realization that the bottleneck in AI compute is no longer chip design or model architecture. It is energy. For founders watching this play out, three implications stand out. First, the nuclear renaissance is real and venture-backed, and it creates opportunities in adjacent layers such as reactor simulation software, nuclear fuel supply chains, and regulatory technology. Second, the partnership with Nvidia signals that the chip giant is actively looking for power solutions, which means any startup solving the energy density problem for AI workloads has a potential anchor customer. Third, Valar's aggressive legal posture toward the NRC shows that regulatory disruption is a legitimate startup strategy in the nuclear space, provided you have the capital and legal team to see it through.
The broader narrative is unmistakable. The AI industry is consuming electricity at a pace that would have seemed absurd five years ago. Data centers already account for roughly 2 percent of global electricity consumption, and that share is climbing fast. Nuclear power, long dismissed as too expensive and too slow, is suddenly the technology that hyperscalers, venture capitalists, and defense contractors are all betting on. Valar Atomics is not the only startup in this race, but it is the one that just proved it can run an AI chip on its own reactor. That single data point is worth more than a thousand slide decks about SMR economics. Whether Valar can scale from a proof-of-concept to hundreds of deployed reactors is an open question that will take years to answer. But the direction of travel is clear: the AI infrastructure buildout is going atomic, and the first mover with working hardware gets to set the pace.
