SSI says its research is ready to scale with NVIDIA

Safe Superintelligence Inc. (SSI), the research lab Ilya Sutskever founded in 2024 to pursue safe superintelligence, announced a long-term partnership with NVIDIA on July 27, 2026. NVIDIA is investing in SSI and giving it access to Vera Rubin systems. SSI says this will increase its compute tenfold over 12 months. The lab says its research is ready to scale but has not disclosed the research.

Ilya Sutskever: “Time to scale that SSI.”

NVIDIA AI Infrastructure: SSI will use Vera Rubin systems.

SSI said NVIDIA was making a substantial investment under a long-term strategic partnership. SSI said the agreement would allow it to increase its compute tenfold during the following 12 months.

In the formal announcement, NVIDIA said the investment and Vera Rubin access would increase SSI's compute by an order of magnitude. The companies said they would collaborate on current and future NVIDIA compute platforms. NVIDIA said it entered the partnership after receiving rare access to SSI's closely guarded research.

The formal announcement describes NVIDIA's investment as substantial but does not disclose an amount.

The Financial Times reports that NVIDIA's investment is $5 billion. Reuters also reports, citing a person briefed on the deal, that NVIDIA will make a $5 billion equity investment in SSI.

Reuters says the companies announced the partnership without disclosing its financial terms. Reuters also reports that the agreement gives SSI access to Vera Rubin hardware and that the deal came together within weeks.

On the Dwarkesh Podcast on November 25, 2025, Sutskever said another 100-fold increase in scale would change AI systems but would not transform everything on its own. At 21:28, he said the field was returning to research using large computers.

At 24:35, Sutskever called models' weak generalization compared with people a fundamental problem. He contrasted the amount of data models need with how reliably people learn newer domains such as mathematics and coding.

At 38:00, he said research requires compute, but not necessarily the most available. SSI had enough to test its ideas; more would help build the best system.

At 42:46, Sutskever said SSI was focused on research. A straight-shot path to superintelligence had merit, he said, but longer timelines and the value of public access could change that plan.

At 46:47, he said even a straight-shot scenario would involve gradual release. The system would learn through deployment rather than arrive with all knowledge. At 56:10, he said he had shifted toward incremental deployment.

At 1:18:13, when asked what SSI planned to do differently, Sutskever pointed to ideas about generalization. He called SSI an “age of research” company, said it had made progress, and did not disclose its approach.

The interview also covers alignment, care for sentient life, limits on the power of the most powerful systems, self-play, multi-agent learning, and research taste. SSI and NVIDIA's July 2026 announcement does not identify which of the interview topics, if any, form SSI's current research direction.