Nvidia Backs Sutskever's SSI, a Lab With No Product Yet

$32 billion. That was the valuation attached to Safe Superintelligence in February 2025, for a company founded the previous year that has never shipped anything. On July 27, Nvidia added its own money to that number.

Bloomberg put the investment at $5 billion; TechCrunch's source described it more loosely as multiple billions. Before this round, SSI had raised roughly $3 billion in total, $1 billion at its 2024 founding and $2 billion in February 2025, from backers including Andreessen Horowitz, Alphabet, Lightspeed, GV and Sequoia.

The terms, as far as they are known

The deal is not a straight cash injection. SSI gets access to Nvidia's Vera Rubin platform, which Ilya Sutskever's lab expects to increase its available compute by an order of magnitude. Nvidia gets something it usually cannot buy: access to SSI's tightly held research, plus collaboration on the design of its current and future compute platforms.

"We have research that is worthy of scaling up, and having access to a big NVIDIA computer will let us do so. We are confident that our big bet on the Vera Rubin platform will take us to the next level." — Ilya Sutskever

Some coverage framed the deal as SSI moving away from Google silicon. That reading deserves a caveat: SSI signed a Google Cloud partnership in April 2025 and Alphabet is an investor, and neither company has announced an exit from that arrangement. What the Nvidia deal clearly signals is a compute realignment, not a confirmed divorce.

Compute as the currency of this deal

The structure is the interesting part. A chipmaker is trading hardware access and capital for research access and design input from a lab with no revenue. That is a different kind of transaction than a venture round, and it is becoming common: compute is now the scarce asset, and the companies that control it can buy positions that money alone would not open.

It also puts SSI in an unusual spot. Sutskever's stated plan has always been to work toward safe superintelligence directly rather than shipping intermediate products. That thesis needs enormous compute and produces no revenue to pay for it, which makes a partner like Nvidia less of an option and more of a requirement.

Our read

For anyone running a business rather than a research lab, the news itself changes nothing this quarter. What it illustrates is worth holding onto: the AI market is increasingly organized around access to compute rather than around products.

That has a practical consequence. When your vendor's roadmap depends on a compute partnership rather than on customer revenue, the vendor's priorities can shift for reasons that have nothing to do with you. It is a reason to keep switching costs low and to prefer providers whose commercial incentives are aligned with shipping to customers, not with proving a thesis.

It is also worth noting what this deal does not include: a product, a date, or a described capability. A $32 billion valuation resting entirely on a research bet is a reasonable venture position. It is not a procurement plan.

Sources: TechCrunch, The Decoder