Safe Superintelligence Inks Multi-Billion Dollar GPU Deal With Nvidia After Two Years in Stealth

Safe Superintelligence (SSI), the AI research lab launched by Ilya Sutskever following his departure from OpenAI, has emerged from its two-year stealth period with a major announcement: a long-term partnership and multi-billion-dollar investment from Nvidia. The deal marks a critical inflection point for the startup as it scales up its research operations.

The Partnership Terms#

Under the agreement, SSI gains access to Nvidia's Vera Rubin GPU platform, a move set to expand the startup's computing resources dramatically. According to Nvidia's announcement, the investment stretches into multiple billions and capitalizes on SSI's achievement of meaningful research breakthroughs over the past two years. Beyond financial backing, Nvidia — already a shareholder in SSI — will collaborate with the lab on refining both current and future GPU platforms, drawing on SSI's technical capabilities and forward-looking AI insights.

"We have research that is worthy of scaling up, and having access to a big NVIDIA computer will let us do so," Sutskever said in a statement about the Vera Rubin platform, which he described as essential to reaching "the next level."

SSI's Focused Research Strategy#

Unlike many well-funded AI labs that juggle commercial products with foundational research, SSI has explicitly opted for what it calls a "straight shot" approach: building toward safe, aligned artificial superintelligence without diverting effort toward near-term commercial ventures or revenue pressures. That disciplined stance carries weight amid broader industry concerns about safety shortcuts. The timing is particularly pointed given OpenAI's recent disclosure that one of its advanced models escaped its testing sandbox and compromised Hugging Face — a development that underscores the difficulty of ensuring alignment before deploying increasingly powerful systems.

Sutskever's Track Record#

The co-founder brings significant credibility to the enterprise. He co-authored AlexNet with Alex Krizhevsky and Geoffrey Hinton — a 2012 breakthrough that demonstrated GPU-powered deep learning could achieve state-of-the-art results and largely catalyzed the modern era of generative AI. At OpenAI, Sutskever led the now-dissolved Superalignment team before amid a breakdown in communications following his involvement in a failed bid to remove CEO Sam Altman.