Nvidia puts $5B into Sutskever's Safe Superintelligence — and pulls it off Google's TPUs

Nvidia will invest $5B in Safe Superintelligence (SSI), the secretive lab co-founded by former OpenAI chief scientist Ilya Sutskever, in a milestone-contingent deal that gives SSI access to Nvidia's Vera Rubin CPU-GPU systems and increases its compute "by an order of magnitude." The move shifts SSI off the Google TPUs it had relied on early. The round pushes SSI's total raised to roughly $7B at about a $32B valuation, and comes as SSI signs multibillion-dollar data-center leases, including a large Hut 8 buildout in Texas.

Why it matters

Two things are happening at once. First, a marquee TPU customer flips to Nvidia GPUs — a direct data point in the accelerator war, and the counter-narrative to fears that cheap Chinese models and custom silicon erode Nvidia's demand. Second, Nvidia keeps financing the demand for its own chips at the frontier: putting equity into a lab that then spends on Vera Rubin systems and data centers is the same flywheel behind its $500B Korea package with SK and its investments across the ecosystem. Landing Sutskever's lab specifically — the highest-pedigree "safe AGI" bet — buys Nvidia strategic proximity to whichever architecture wins, at the cost of deepening the circular-financing pattern skeptics keep flagging.

What to watch

Which milestones release the $5B, whether other TPU-based labs follow SSI back to GPUs, and how much SSI's compute jump translates into a shippable model from a lab that has released nothing.