Nvidia Commits $5 Billion to Ilya Sutskever’s Safe Superintelligence

Safe Superintelligence Inc., the AI research lab led by OpenAI co-founder Ilya Sutskever, has landed a roughly $5 billion investment from Nvidia, according to a joint statement from the two companies. The deal pairs Nvidia’s capital with early access to its next-generation Vera Rubin computing platform, giving the famously secretive startup a direct line to the chipmaker’s most advanced hardware before it reaches the broader market. Terms beyond the investment amount were not disclosed.

A Valuation That Has Grown Sixfold in Under a Year

Safe Superintelligence has taken an unusual path for a company its size: it has raised billions of dollars without shipping a public product. The lab’s first round in September 2024 valued it at $5 billion. By early 2025, a $2 billion raise pushed that figure to roughly $32 billion, backed by Greenoaks, Andreessen Horowitz, Lightspeed Venture Partners, DST Global, and both Alphabet and Nvidia.

Google Cloud has since become a major infrastructure provider for the lab’s training workloads, similar to how Together AI’s recent $800 million Series C underscored how much capital is flowing into the compute layer beneath frontier AI labs. Nvidia’s new commitment adds to a cap table that already reads like a who’s-who of AI investors, and signals that Nvidia sees strategic value in supplying compute directly to one of the field’s best-funded research efforts rather than just selling chips at arm’s length.

From OpenAI’s Chief Scientist to SSI’s CEO

Sutskever left OpenAI in May 2024 to start Safe Superintelligence with Daniel Gross and Daniel Levy, positioning the lab around a single stated goal: building safe superintelligent AI systems without the commercial pressure of shipping consumer products along the way. The company’s leadership shifted in mid-2025 when Meta recruited Gross, then SSI’s chief executive, to help run its new superintelligence lab after Meta’s attempt to acquire SSI outright was turned down. Sutskever stepped into the CEO role himself, with Levy remaining as a co-founder and steering the technical work.

Why Nvidia Is Backing Safe Superintelligence

For Nvidia, the investment extends a pattern of putting capital behind the AI labs most likely to consume large volumes of its highest-end chips. Reports around the deal describe Safe Superintelligence shifting a portion of its training infrastructure toward Nvidia GPUs, a move that would reduce its reliance on Google’s TPUs even as Google remains both an investor and cloud partner.

That shift echoes what’s happening across the AI infrastructure market more broadly — Fluidstack’s $830 million Series A for GPU cloud infrastructure was raised on the same bet, that whoever controls compute access controls the pace of AI progress. Tying a research lab’s roadmap to Nvidia’s own hardware release schedule, through early access to Vera Rubin, gives Nvidia visibility into how its next platform performs on frontier-scale training runs well before general availability. It’s the same compute-first logic driving California’s AI startup boom, where funding increasingly follows access to chips as much as it follows research talent.

What It Means for the Broader AI Race

The Nvidia deal cements Safe Superintelligence as one of the best-capitalized AI labs in the world despite having released nothing to the public. For a company built around the premise that safety research should happen away from commercial pressure, that kind of funding cushion matters: it buys Sutskever’s team time and compute without the need to ship a product to cover payroll. Whether that approach produces a genuine breakthrough, or simply the best-funded research group in AI history, is the open question investors are betting billions on.