Summary
Safe Superintelligence (SSI), the secretive AI lab founded by former OpenAI chief scientist Ilya Sutskever, has announced a long-term strategic partnership with Nvidia that includes a reported $5 billion investment and a commitment to increase SSI’s compute capacity by 10x over the next 12 months using Nvidia’s next-generation systems. The announcement is notable for a single phrase SSI used to justify the deal: their research has reached the point where it is “worth scaling” — a statement that carries significant weight given Sutskever’s public skepticism about simply throwing more compute at existing architectures.
This video walks through nine key details of the deal, then rewinds to trace Sutskever’s intellectual journey: his December 2024 NeurIPS prediction that pre-training as currently practiced would eventually hit a ceiling, and his November 2025 interview with Dwarkesh Patel where he described a shift from the “age of scaling” back to an “age of research with big computers.” Crucially, Sutskever later clarified he was not saying scaling is dead — only that something important would still be missing without a new underlying research breakthrough.
The core question the video explores is what SSI may have actually discovered that convinced Nvidia to grant the company rare access to its closely guarded internal systems before signing. With SSI operating as a pure research organization — no API business, no enterprise contracts, no inference costs — the partnership signals that at least one major hardware company believes Sutskever’s team may have found the next architecture worth betting on at scale.
📺 Source: TheAIGRID · Published July 29, 2026
🏷️ Format: News Analysis
