Safe Superintelligence is valued at $30 billion

Background

Safe Superintelligence Inc. (SSI) was founded by Ilya Sutskever together with Daniel Gross and Daniel Levy.^[2] By the time of the events described here the company had already raised $1 billion.^[2] In September 2024 it was valued at $5 billion.^[1]

The March 2025 round

In March 2025 SSI closed a funding round led by Greenoaks Capital that valued the company at $30 billion.^[1] That was six times the $5 billion valuation it had received in September 2024.^[1] The figure drew attention for a second reason: at the time of the round the company had no revenue and about 20 employees.^[1]

A valuation is not a measure of current earnings but of what investors expect a company to be worth later, which is why very young research labs can carry very large numbers. What the March 2025 round showed was that a small team with no sales could still be priced alongside much larger and older companies, on the strength of who was building it and what it said it was aiming at.

The April 2025 report

On 12 April 2025 TechCrunch, citing the Financial Times, reported that SSI had raised an additional $2 billion at a $32 billion valuation, in a round reportedly led by Greenoaks.^[2] The same report said the company had already raised $1 billion, and that SSI did not comment on the new funding.^[2]

The two accounts do not match exactly: one puts the round at a $30 billion valuation, the other at $32 billion, and the April report describes a further $2 billion on top of money already raised. SSI did not comment on the reported new funding.^[2] Reports of this kind often move faster than confirmation, and the numbers attached to a private company can differ between outlets before anything is officially disclosed.

Compute and an approach from Meta

Also in April 2025, Google Cloud agreed to supply TPUs for SSI's research.^[1] TPUs are chips designed specifically for machine learning; access to them, rather than to general-purpose hardware alone, determines how quickly a lab can run large training experiments. For a company of about 20 people, an agreement of this kind is a large part of its practical capacity.

In the first half of 2025 Meta attempted to acquire SSI, and Sutskever rebuffed the approach.^[1] Acquisitions of research labs are one way large technology companies absorb scarce expertise; a refusal keeps the lab independent and leaves its direction in the hands of its founders.

A later Nvidia partnership

In July 2026 SSI announced a long-term partnership with Nvidia giving it access to the Vera Rubin GPU platform, which is expected to increase its compute resources by an order of magnitude.^[3] Nvidia's investment in that partnership stretched into multiple billions, with Bloomberg reporting a $5 billion figure.^[3]

Why it matters

The sequence is a compact picture of how frontier AI work was financed in this period. A lab with no revenue and a few dozen people could raise money at valuations in the tens of billions; chip suppliers became partners and investors rather than mere vendors; and the largest platform companies were willing to buy such labs outright. The sums involved were directed less at products already on the market than at the compute and people needed to attempt something that did not yet exist. Whether that bet pays off is a separate question, and one that the valuations themselves do not answer.