Ilya Sutskever's Safe Superintelligence (SSI) partnered with Nvidia to scale its compute 10x. Nvidia also invested in SSI, marking the chipmaker's second known direct investment in a pure AI safety lab.
Key facts
- SSI partnered with Nvidia to scale compute 10x.
- Nvidia invested in SSI; amount undisclosed.
- SSI founded June 2024 by Ilya Sutskever.
- Nvidia also invested in Inflection AI and CoreWeave.
- SSI has not disclosed valuation, revenue, or compute capacity.
Ilya Sutskever's Safe Superintelligence (SSI) partnered with Nvidia to scale its compute 10x. Nvidia also invested in SSI, according to a post on X by @rohanpaul_ai Source: @rohanpaul_ai. The announcement did not disclose the investment size, SSI's current compute capacity, or the timeline for the 10x expansion.
SSI was founded in June 2024 by Sutskever, former chief scientist at OpenAI, along with Daniel Gross and Daniel Levy. The startup's stated mission is to build safe superintelligence, a goal that has drawn skepticism from some researchers who argue that safety and capability scaling are in tension. The partnership with Nvidia suggests SSI is pursuing the same compute-intensive path as competitors like OpenAI and Anthropic, but with a safety-first framing.
Nvidia's investment in SSI follows a pattern of the chipmaker making strategic equity bets in AI labs, including earlier investments in Inflection AI and CoreWeave. By tying compute supply to equity, Nvidia locks in demand for its GPUs while gaining insight into frontier research directions. The deal also signals that Nvidia is betting on multiple frontier labs to sustain demand for its H100 and B200 GPUs, rather than relying solely on hyperscalers.
The 10x compute target raises questions about SSI's capital needs. For context, training a single frontier model at scale can cost tens of millions of dollars in compute. A 10x increase from an undisclosed baseline is difficult to assess, but if SSI is starting from a modest cluster, the expansion may still leave it far behind the compute budgets of OpenAI or Anthropic, which each operate hundreds of thousands of GPUs.
SSI has not disclosed its valuation, revenue, or compute capacity, making the scale of the 10x target difficult to assess. The startup operates with a small team and has not released any public demos or benchmarks, according to its website. Sutskever's safety-first approach may limit SSI's commercial attractiveness, but the Nvidia tie-up provides capital and hardware access that could accelerate its research timeline.
Key Takeaways
- SSI partners with Nvidia for 10x compute scale; Nvidia also invests.
- Details on investment size and timeline undisclosed, raising questions about the startup's capital needs.
What to watch
Watch for SSI's first public benchmark release or model demo, which would validate whether the 10x compute expansion translates to tangible safety research outputs. Also track Nvidia's Q1 2026 earnings call for any mention of the investment size or strategic rationale.
[Updated 28 Jul via dcd_news]
Nvidia's investment in SSI is valued at $5 billion, according to a report from DatacenterDynamics, making it Nvidia's largest known direct investment in an AI lab. Prior to the deal, SSI had been using Google TPUs for its compute needs, signaling a strategic shift to Nvidia's H100 and B200 GPUs [per DatacenterDynamics].
