Nuclear startup Valar Atomics in talks to raise new funding at $6B valuation

Valar Atomics wants $6 billion for a reactor design that has never produced grid power. Three years old, zero commercial units, a single demonstration feeding an Nvidia chip — and Sequoia is reportedly ready to lead a $1 billion equity round at a valuation that tripled in six months. That is not a funding round. It is a marker.

The Information broke the talks. TechCrunch confirmed the structure: $450 million already in the door at $2 billion, split between equity and debt. Now new money at $6 billion. The same company, different price tags, same cap table. This is how AI-era fundraising works — staggered closings, rising marks, paper gains that look like traction. Outsiders see a $6 billion startup. Insiders see a series of option exercises.

Valar's pitch is timing, not technology. Data centers are starving for electrons. Utilities cannot build fast enough. Nuclear, long written off as too slow and too expensive, suddenly looks like the only supplement that scales. The startup's helium-cooled, high-temperature gas reactor is theoretically factory-buildable, theoretically cheaper, theoretically faster to license. Theoretical is doing a lot of work in that sentence.

NuScale Power holds the only U.S. design approval for an SMR. Kairos Power and TerraPower — backed by Bill Gates — are chasing the same industrial customers. Valar has Palmer Luckey and Palantir's CTO on its cap table. That is a defense roster, not a construction crew. The company promises hundreds of reactors. It has not broken ground on one.

The Nvidia partnership is the shrewdest move Valar has made. A photo of a reactor lighting a GPU chip travels farther in Silicon Valley than a thousand pages of regulatory filings. It frames nuclear as compute infrastructure. That framing attracts Sequoia. It also obscures the hard path: the Nuclear Regulatory Commission does not care about AI timelines. Valar sued the NRC last year alongside other developers, challenging the licensing framework itself. Aggressive litigation is a strategy, but it is not a schedule.

Sequoia knows the risk. The firm has backed hard tech before — SpaceX, Stripe, Airbnb. But those bets faced market risk, not physics and administrative law intertwined. A helium-cooled gas reactor operates at temperatures that stress materials science. Factory fabrication demands supply chains that do not exist. Fuel qualification takes years. The gap between "small amount of power to an Nvidia chip" and "hundreds of SMRs powering data centers" is measured in decades, not funding rounds.

The $6 billion valuation prices in certainty that does not exist. It prices in a regulatory surrender that has not happened. It prices in a manufacturing scale-up that has never been attempted for this reactor type. Sequoia is not buying a power plant. It is buying a call option on the energy transition compressing into the AI buildout.

That bet may pay. The demand vacuum is real. Gas peakers cannot fill it fast enough. Renewables cannot firm it alone. Nuclear is the only baseload that scales without carbon. But the winner will be the company that clears the NRC first, not the one with the flashiest demo. NuScale already cleared that bar. Valar is still running at it.

The multi-tranche structure protects early investors. They got in at $2 billion. New money enters at $6 billion. If Valar stalls, the early money still marks up. If it succeeds, everyone claims foresight. The structure rewards momentum over milestones. That is the logic of the current cycle — and its danger.

Valar Atomics may yet build the reactor that powers the next generation of AI. But today it owns a demonstration, a lawsuit, and a valuation that assumes the hardest problems are already solved. They are not. The reactor does not care about cap tables. The NRC does not care about Sequoia. Physics and procedure move at their own pace. Capital rarely waits.