Nvidia and Safe Superintelligence Inc. (SSI), the AI lab founded by former OpenAI chief scientist Ilya Sutskever, announced a long-term strategic partnership Monday under which Nvidia is investing in SSI and giving it early access to its next-generation Vera Rubin chip platform. Bloomberg and other outlets, citing people familiar with the deal, put Nvidia’s investment at $5 billion — one of the chipmaker’s largest single bets on an AI lab to date.
Nvidia said the Vera Rubin access will expand SSI’s available compute “by an order of magnitude” over the coming year.
“We have research that is worthy of scaling up, and having access to a big Nvidia computer will let us do so,” Sutskever said in a joint statement. Nvidia CEO Jensen Huang credited Sutskever with “fundamental breakthroughs at the foundation of modern AI,” pointing to his earlier work on AlexNet and sequence-to-sequence learning, and the research that led to OpenAI’s o1 reasoning models.
A quiet, product-less startup
SSI has published no research and shipped no product since Sutskever launched it in 2024 alongside former Apple AI lead Daniel Gross and researcher Daniel Levy. The company has instead pursued what Sutskever calls a “straight shot” toward safe superintelligence, avoiding the commercial release cycle that drives rivals such as OpenAI and Anthropic.
That approach hasn’t discouraged investors. SSI launched at a $5 billion valuation with $1 billion raised, then reached a $32 billion valuation last year as Alphabet, Nvidia and other backers including Andreessen Horowitz, DST Global, Greenoaks and Sequoia Capital pushed its total funding past $3 billion. The new investment adds to a string of multibillion-dollar infrastructure deals Nvidia has struck with AI labs this year as it works to lock in demand for its GPUs.
Part of a broader compute race
The SSI deal lands as AI capital spending across the industry keeps climbing, with labs racing to secure chips and power years ahead of need. Unlike most well-funded AI startups, SSI has given no timeline for a public product, betting instead that more compute now will let a small team reach a safe superintelligent system before it needs to compete on revenue.