NVIDIA partners with Sutskever's Safe Superintelligence to scale its research

NVIDIA's compute partnership with Safe Superintelligence gives the chipmaker a foothold inside one of the industry's most secretive labs. SSI, co-founded by former OpenAI chief scientist Ilya Sutskever, has famously shipped no products and disclosed almost nothing about its research. The deal scales SSI's work on NVIDIA's next-generation Vera Rubin platform and, per TechCrunch, grants NVIDIA 'rare access' to the lab — a valuable window given SSI's opacity.

The mechanics follow NVIDIA's now-familiar playbook: provide compute scale (and, as an existing investor, capital) to a frontier lab in exchange for strategic access and platform lock-in. It mirrors NVIDIA's simultaneous $500B+ SK Group deal and its role anchoring national AI infrastructure, part of a week defined by gigawatt-scale compute commitments.

The partnership drew commentary on the contradictions it embodies. NVIDIA is now backing both OpenAI and the 'ultra-secretive' SSI while simultaneously championing open-weight models through the new Open Secure AI Alliance — a 'super-closed' lab and an open-model coalition under the same corporate umbrella. Observers noted the one-directional compute dependencies this creates, with NVIDIA at the center of nearly every frontier effort regardless of philosophy.

Skeptics question what SSI actually offers in return beyond prestige, given it has produced no public results, and whether 'safe superintelligence' remains a fundable thesis absent products or revenue. The deal also raises governance questions about how much visibility any single hardware vendor should have across competing frontier labs. What to watch: whether SSI ever discloses research output, and how NVIDIA balances its open and closed bets.