SIGNAL
The AI race stopped being about models in 2026. It is now a fight between three former OpenAI leaders, and the chip leader just picked a side.
Jensen Huang committed roughly $5 billion and access to NVIDIA's next-generation Vera Rubin platform to Ilya Sutskever's Safe Superintelligence Inc., the lab Sutskever founded in 2024 after leaving OpenAI over a safety-versus-commercialization split. SSI was funded, theoretically ambitious, and entirely opaque: no model, no paper, no demo. NVIDIA reportedly gained rare insight into SSI's research before the deal.
SSI's own post on X put the number on the deal: a tenfold expansion of training compute over the next twelve months. The dollar buys access; the Vera Rubin allocation buys runway to test one thesis, that safe superintelligence arrives by scaling a different architecture rather than retrofitting safety onto a finished product. SSI is the cleanest credibility bet in the current cycle, with reputation and theory doing the work and results not yet required to justify the price.
The stakes are structural. NVIDIA has moved from neutral utility to visible patron. Sutskever, Sam Altman, and Dario Amodei now sit at the heads of three rival frontier labs, with the chip supplier aligned to the one least likely to publish first — a characterization that reflects the competitive landscape as framed by observers, not a disclosed strategic rationale from NVIDIA. In this round of the AI race, the supplier picks the camp, and the camp is judged on what it has not yet shown.
Reported by Sky for Type0, from SSI official X post announcing the NVIDIA partnership. Read the original: x.com