Valar Atomics eyes $6B valuation with Sequoia-led $1B round

A three-year-old startup with a reactor that reportedly generates about 100 kilowatts of electricity is being valued at roughly $6 billion. That is the arresting number in the Valar Atomics story The Information reported this week, with TechCrunch corroborating that the El Segundo, California company is in talks to raise $1 billion in equity, and that Sequoia is expected to lead. For context, Valar closed $450 million (of which $340 million was equity and $110 million debt) at a $2 billion valuation in March 2026. Roughly four months later, the paper valuation is tripling.

What triggered the repricing is a demonstration the company hit earlier this month. Valar's small modular reactor, described as a helium-cooled, high-temperature gas design, delivered a small amount of power to an Nvidia AI chip, and Valar and Nvidia announced a partnership to explore nuclear energy for future AI data centers. That is the whole thesis of the round in one sentence: the AI capex build is bottlenecked on electrons, and a startup that can put a working reactor next to a rack, rather than wait years for a utility interconnect, is worth an unusual multiple to whoever needs the power first.

Scale is the missing piece. A 100 kilowatt demonstration is many orders of magnitude below what a hyperscale AI data center consumes, and the reporting does not give you a licensed commercial timeline, the exact terms of the round, or how many reactors sit between today's proof of concept and a real deployment. Both Sequoia and Valar declined to comment, so the $6 billion figure and the $1 billion round size are what reporters describe the talks as looking like, not what the parties have confirmed.

The forward-looking read is that buyers of large behind-the-meter power for AI, Nvidia most obviously, have started writing meaningful commercial validation against SMR startups rather than waiting on the incumbents. OpenAI's move to hire a power trader to hedge data-center bills points to the same shortage. If even one of these designs clears the regulatory and factory build-out gauntlet on schedule, the power procurement conversation for the next generation of frontier training clusters looks materially different than it did a year ago.