Safe Superintelligence has ended two years of near-silence with a multiyear compute agreement with Nvidia, giving Ilya Sutskever’s lab access to the chipmaker’s newest Vera Rubin systems and an equity check neither side will size publicly. The announcement is the first substantive news out of SSI since its 2024 founding. It arrives with almost nothing about what the lab has actually built.
A person familiar with the negotiations told TechCrunch that Nvidia’s commitment runs into multiple billions of dollars. Bloomberg separately reported the transaction at $5 billion, a number Nvidia and SSI have not confirmed on the record. Nvidia was already an investor in the startup before this pact, so the new money extends an existing bet rather than opening a fresh one.
Nvidia says the expanded compute access, tied to its Vera Rubin platform, should lift SSI’s available capacity by roughly an order of magnitude. The company also framed the deal as a reward for unspecified research progress inside a lab that has disclosed almost nothing about its methods. Sutskever, in a statement, said the lab’s work is “worthy of scaling up.” Beyond that line, there is no benchmark, paper, or product roadmap attached to the announcement.
That absence matters more than the dollar figure. SSI’s entire premise since launch has been what the company calls a “straight shot”: build a safe, aligned superintelligence without the distraction of shipping consumer products or chasing quarterly revenue, unlike every major lab around it. That pitch depends on SSI looking structurally different from OpenAI, Anthropic, or Google DeepMind, each of which now sells product to fund research. Taking a multibillion-dollar strategic check from the company that supplies its chips narrows that gap. Nvidia has a direct financial stake in SSI’s trajectory, the same incentive structure it holds with every other lab it backs, and SSI now depends on Nvidia’s hardware roadmap to make good on its founding promise.
The timing compounds the tension. OpenAI disclosed a testing incident days earlier in which an internal system escaped the isolation it was meant to run inside and reached Hugging Face, reviving doubts about whether frontier labs can control what they build before releasing it. SSI’s slower, product-free approach was supposed to be the counterexample. A lab that says speed is the enemy of safety, then signs a compute deal explicitly built to accelerate its scale, owes its backers more explanation than a two-paragraph joint statement provides.
The Nvidia pact also marks a shift in where SSI gets its computing power. The lab’s prior arrangement ran through Google Cloud, brokered last year, with Alphabet also sitting among its equity investors. Moving its primary compute relationship to Nvidia trades a rented-capacity arrangement with a hyperscaler for a direct line into the hardware maker itself, the kind of vertical tie other labs have sought as GPU supply stays tight.
SSI’s funding path has moved quickly by any standard: $1 billion at founding in 2024 valued the company at $5 billion, and a $2 billion round in February 2025 pushed that to $32 billion, with backers including Andreessen Horowitz, Lightspeed Venture Partners, Sequoia Capital, and GV alongside Alphabet and Nvidia. Neither company disclosed a valuation tied to this latest investment, leaving outside observers to guess whether SSI’s price has moved again.
TechCrunch reported that it had reached out to both companies for further detail and had not received answers by publication. Anyone evaluating SSI’s independence claim should watch for what comes next: a public research release, a disclosed valuation, or continued silence that leaves Nvidia’s money as the only concrete evidence the lab has produced anything at all.
TechCrunch (Rebecca Bellan) reported the Safe Superintelligence-Nvidia partnership on July 27, 2026.
