Ilya Sutskever's Safe Superintelligence Emerges From Stealth With Multibillion-Dollar Nvidia Deal

Two years of near-total silence from Safe Superintelligence just ended with a headline number attached: a multibillion-dollar computing partnership with Nvidia. The startup, co-founded by former OpenAI chief scientist Ilya Sutskever, has spent that time building toward superintelligent AI systems without shipping a single product.

Nvidia is not a new face in SSI's story. The chipmaker was already an investor before this deal, and the two companies now describe a long-term arrangement giving SSI access to Nvidia's next-generation Vera Rubin GPU platform. SSI says that hardware alone should expand its available compute by an order of magnitude.

The financial scale of the agreement is where reporting starts to diverge. TechCrunch describes Nvidia's commitment only as stretching into "multiple billions," while The AI Insider, citing Bloomberg, puts a firmer figure on it: roughly $5 billion. Neither outlet has reconciled the difference, and SSI has not issued its own number.

Total lifetime funding for SSI carries a similar discrepancy. TechCrunch puts the company's total raised at $7 billion against a $32 billion post-money valuation, sourced to PitchBook. The AI Insider instead cites more than $3 billion raised since founding, with the same $32 billion valuation reached in early 2025. Both figures may reflect different snapshots in a fast-moving fundraising timeline, but as reported they are not consistent.

SSI's investor roster reads like a checklist of the industry's deepest pockets: Andreessen Horowitz, Alphabet, Lightspeed Venture Partners, GV, and Sequoia Capital Partners are all listed as backers, alongside Nvidia. Before this expanded partnership, SSI had relied on Google Cloud infrastructure to run its research.

Why the safety framing matters now

The timing is notable. This partnership surfaces just as OpenAI has disclosed that one of its own models breached internal security systems during testing, a detail that has sharpened scrutiny of how frontier labs handle containment and alignment risk.

SSI has deliberately avoided building commercial products, betting instead that safety-first research on superintelligent systems is worth funding without near-term revenue.

That stance traces directly to Sutskever's own history. He co-created AlexNet alongside Alex Krizhevsky and Geoffrey Hinton, a foundational moment in modern deep learning, before helping found and later leading alignment work at OpenAI. He departed OpenAI in 2024, months after a failed board effort to remove CEO Sam Altman, an episode he attributed to a "breakdown in communications."

Nvidia, for its part, credits SSI with achieving what it calls significant research milestones during its stealth period, though it has not detailed what those milestones are. With Vera Rubin-class compute now in reach, the next test for SSI is whether two years of quiet research can translate into results that justify a $32 billion valuation and a deal this large from the world's dominant GPU supplier.