Nvidia made a reported $5 billion bet on Sutskever's Safe Superintelligence

Nvidia is investing in Safe Superintelligence, the AI startup founded by former OpenAI chief scientist Ilya Sutskever. The deal was announced on July 27, 2026. Nvidia’s announcement gives no amount; Bloomberg reported the deal size at $5 billion, and a source told TechCrunch the investment stretches into multiple billions.

What the partnership includes

The investment comes bundled with a deep infrastructure relationship:

• SSI gets access to Nvidia’s next-generation Vera Rubin platform
• The platform is expected to increase the startup’s compute by an order of magnitude
• SSI previously raised $3 billion, most recently at a $32 billion valuation in February 2025

The compute access is the quiet win inside the bigger headline. SSI partnered with Google Cloud last year to power its research, and the announcement does not say whether that changes. Either way, one of the most closely watched research labs in the world now has Nvidia’s next platform in its plans, and Nvidia paid to put it there.

A $32 billion valuation with no product

The wild part: two years after launch, Safe Superintelligence still has no commercial product. That is not an oversight but the company’s stated strategy. SSI calls it a “straight shot” to safe superintelligence: no commercial releases, no distractions, just research.

Investors are effectively pricing a single outcome. Either SSI reaches its goal and the valuation looks quaint, or it does not and the billions funded a research program. There is no middle case with products and revenue to fall back on.

Nvidia at the center of everything

Zoom out and a pattern completes. Nvidia now backs OpenAI’s data centers, invests in SSI, and sits at the center of the entire AI economy as chipmaker, lender and investor at once.

Every major bet in the AI race now runs through the same company, which supplies the chips, finances the buyers and holds equity in the labs. Whether a $32 billion valuation with zero product is genius or a bubble signal may matter less than the fact that Nvidia wins in either scenario, as long as the research keeps demanding GPUs.