Nvidia SSI investment moves Sutskever’s lab toward Nvidia chips

Nvidia’s SSI investment gives the chipmaker a stake in one of the most closely watched frontier AI labs and gives Safe Superintelligence access to Nvidia’s next-generation Vera Rubin GPU platform. Nvidia and SSI announced the long-term partnership Monday, but did not disclose the size of the investment beyond calling it “substantial.”

The deal matters because SSI, founded by former OpenAI chief scientist Ilya Sutskever, had mainly relied on Google’s TPU chips, according to The Wall Street Journal. A shift toward Nvidia hardware gives Nvidia another high-profile customer in frontier AI research and pushes back against Google’s effort to make TPUs a stronger alternative to Nvidia GPUs during the AI buildout.

What is Nvidia investing in SSI for?

Under the partnership, SSI will use Nvidia’s Vera Rubin GPU platform, which the companies said will increase the lab’s compute capacity by a factor of ten. They did not give a timeline for the rollout, the number of chips involved or the commercial terms attached to the hardware access.

SSI was launched in 2024 with one stated objective: building safe superintelligence through what it has described as a direct research push. Sutskever has said the company is studying “overlooked aspects of how the human brain functions,” a broad research claim that the company has not translated into disclosed products, revenue or customer metrics.

The company has already raised about $2 billion from investors including Andreessen Horowitz, Sequoia Capital, DST Global and Greenoaks. Its most recent reported valuation was roughly $30 billion. That makes SSI one of the highest-valued private AI labs despite limited public detail about its technical progress.

Why the chip supplier shift matters

Frontier AI labs are constrained by access to compute, and the supplier relationship can shape cost, availability and technical roadmaps. Nvidia’s position has been strengthened by demand for GPU clusters used to train and serve large AI models. Google’s TPUs remain one of the most credible in-house alternatives, particularly for companies willing to build around Google’s stack.

For Nvidia, SSI is a strategic account as much as a financial investment. The company is not just selling accelerators into the AI boom; it is tying leading model developers to its future platforms before those labs standardize around competing chips. Vera Rubin is Nvidia’s next platform after its current generation, and access to it is being positioned by the companies as a way for SSI to scale its research capacity.

For SSI, the arrangement provides a path to more compute without disclosing how much capital the company is spending on infrastructure. The lab’s backers have already financed it at a level usually reserved for companies with visible commercial traction. The new Nvidia partnership gives SSI another resource for its research plan, while leaving unanswered how it measures progress toward its safety and superintelligence goals.

The companies announced the partnership through Nvidia. The Wall Street Journal reported SSI’s prior reliance on Google TPUs.

This story draws on original reporting from The Decoder.