Ilya Sutskever’s Safe Superintelligence and Nvidia Announce Long-Term Strategic Partnership

Comments

Founded in 2024 and led by Ilya Sutskever and Daniel Levy, Safe Superintelligence describes itself as "the world's first straight-shot SSI lab", built around a single mission and ultimately a single product: safe superintelligence. The company deliberately positions itself outside the traditional commercial cycle, with no product deadlines or revenue targets, allowing capabilities and safety to advance "in tandem". Its investors include Andreessen Horowitz, DST Global, Greenoaks, and Sequoia Capital. SSI’s first funding round in September 2024 valued the company at $5 billion, while a $2 billion round in April 2025 lifted its valuation to $32 billion. According to PitchBook estimates, total funding has now reached roughly $7 billion. What makes these figures even more remarkable is SSI's profile: a lab of only around 33 people, with no public product and no published research papers so far.

Neither company disclosed the exact terms of the latest deal, but Bloomberg and other sources estimate Nvidia’s investment at approximately $5 billion. The agreement would increase SSI’s available compute capacity by roughly tenfold and represents a significant evolution from its initial reliance on Google TPUs. Google Cloud remains part of SSI's infrastructure strategy, continuing to provide TPUs alongside the newly expanded access to Nvidia GPUs.

The deal confirms a pattern we have already observed several times across the AI infrastructure landscape.

From Nvidia's perspective, the strategy is becoming increasingly clear: take equity positions in promising frontier AI labs, including SSI and Mira Murati's Thinking Machines Lab, among others, while simultaneously bringing them into its silicon and infrastructure roadmap. This approach secures customer commitment and loyalty long before these companies become major revenue-generating customers. Seen from another angle, these transactions look less like conventional venture investments and more like compute-supply agreements wrapped in strategic equity. SSI gains privileged access to the processors expected to define frontier-model training through the late 2020s, while Nvidia gains a massive-scale environment in which technologies such as Vera Rubin can be deployed, validated, and optimized.

It also raises the familiar question of circularity, or, depending on your perspective, whether this should instead be called a virtuous circle. We have highlighted this self-reinforcing structure across several recent AI mega-deals: compute attracts equity, equity finances additional compute, and that additional infrastructure further increases the value and ambitions of the AI company involved. Similar concerns have emerged around other financing structures in the industry, including arrangements associated with Anthropic and large infrastructure investors. A $32 billion valuation for a company with no shipped product and no revenue represents perhaps one of the most striking examples yet of research credibility, talent, and future technological potential becoming an asset class of their own.

Finally, there is the infrastructure impact. Deals of this magnitude inevitably translate into physical capacity. Training frontier models at this scale requires enormous amounts of power, accelerators, networking, memory, storage, cooling, and data-center infrastructure. SSI's expansion therefore points toward another wave of multi-gigawatt AI compute campuses, reinforcing the extraordinary infrastructure build-out already underway across the industry.