Safe Superintelligence is becoming a central customer in the next phase of AI compute without shipping a public product. On July 27, 2026, NVIDIA said it had entered a long-term strategic partnership with Safe Superintelligence Inc. and made an investment in the company. The concrete number is compute scale: NVIDIA says access to its Vera Rubin platform will let SSI increase its compute by an order of magnitude.

SSI is led by Ilya Sutskever and Daniel Levy and was founded in 2024 to focus on one goal: safe superintelligence. NVIDIA’s release highlights Sutskever’s role in AlexNet, AlphaGo, sequence-to-sequence learning, GPT models, and the research line that led to OpenAI o1. That background helps explain why a lab with little public output can still command scarce infrastructure and investor attention.

The structure of the deal is more revealing than the investment alone. NVIDIA says it entered the partnership after obtaining rare access to SSI’s closely guarded research direction. SSI gets a larger NVIDIA computer based on the Vera Rubin platform. NVIDIA gets a closer technical relationship with a frontier research lab and says the two companies will collaborate on current and future compute platforms using SSI’s view of where AI is heading. The release does not disclose the investment amount, number of systems, power commitment, or delivery schedule.

Vera Rubin is NVIDIA’s upcoming platform for large-scale AI systems, and early access is becoming a strategic asset in itself. For labs chasing frontier capabilities, compute is no longer just a cloud procurement line item. It is a negotiated relationship involving capital, hardware priority, and technical feedback that can shape future accelerator design.

The next test is external evidence. SSI says it has research worth scaling, but the public still has few benchmarks, products, or papers to evaluate. A 10x compute increase raises expectations and pressure at the same time. For NVIDIA, the deal also shows how chipmakers are moving deeper into the research loop: not just selling GPUs to model labs, but using frontier labs as signals for what the next machines need to become.

#ssi

SSI gets 10x more compute as NVIDIA puts Vera Rubin behind Sutskever’s lab