NVIDIA has entered a long-term strategic partnership with Safe Superintelligence, the AI company founded by former OpenAI chief scientist Ilya Sutskever. The deal includes a major investment - reported by Bloomberg at roughly $5 billion - and gives the startup access to NVIDIA's next-generation Vera Rubin computing platform. The partnership accelerates Safe Superintelligence's tightly controlled research while giving NVIDIA deeper insight into the future of AI architecture.
Who is Safe Superintelligence?
Safe Superintelligence was founded by Sutskever alongside Daniel Gross, a former Apple AI lead and co-founder of search engine Cue, and Daniel Levy, a former OpenAI engineer. The company has operated largely in stealth, focusing on building powerful AI systems with safety as a core requirement. Since its founding, Safe Superintelligence has raised $7 billion and reached a valuation of $32 billion, according to PitchBook data. That's a sharp rise from the $5 billion valuation it carried just three months after its September 2024 launch.
The partnership is structured around two-way access. Safe Superintelligence gets the computing power it needs to scale its research on the Vera Rubin platform. NVIDIA, in turn, gains access to the startup's research insights to help shape the design of its current and future computing platforms. Jensen Huang, NVIDIA's founder and CEO, framed the relationship as a bet on Sutskever's track record of fundamental AI breakthroughs. "Ilya has spearheaded groundbreaking innovations that form the foundation of modern AI, starting with AlexNet," Huang said. "We are very excited to see what new breakthroughs Safe Superintelligence will uncover by leveraging NVIDIA's Vera Rubin platform."
Investment scale and market context
While NVIDIA and Safe Superintelligence did not disclose the exact investment amount, multiple outlets have reported the figure. Bloomberg and Reuters each cited sources putting the deal at $5 billion. TechCrunch described the investment as worth several billion dollars. The funds come as the AI hardware market races to support next-generation training workloads, and as NVIDIA's Vera Rubin NVL72 rack - promising 2400 TFLOPS in FP64 precision - enters the spotlight.
Sutskever said the computing access was decisive. "Safe Superintelligence has research that deserves to scale, and having access to NVIDIA's large-scale computers makes that possible," he said. "We are very proud to be partnering with Jensen and the NVIDIA team, and we are confident that this significant investment in the Vera Rubin platform will take us to the next level."
What the deal signals about AI infrastructure strategy
The partnership reverses the usual dynamic where chipmakers sell hardware and stay arms-length from the research labs. Here, NVIDIA is betting capital and silicon on a single, highly focused team. That concentration of resources - and the direct feedback loop between Sutskever's work and NVIDIA's chip roadmap - can compress the time it takes to move from architectural insight to shipping silicon. For executives tracking AI infrastructure, the deal underscores that the largest AI platform providers are now selecting research partners not just as customers, but as co-designers.
The structure also highlights the growing importance of controlled research environments. Safe Superintelligence's research is described by NVIDIA as "tightly controlled," and the partnership was formed only after NVIDIA gained "valuable access" to that work. This suggests that due diligence on AI safety and technical direction is becoming a prerequisite for strategic hardware investments.
Why this matters for executives
When a chipmaker of NVIDIA's scale invests $5 billion in a single AI startup, it signals where the next generation of computing demand is heading. This deal is not just about funding a company - it's about locking in architectural direction for the Vera Rubin platform and beyond. For CEOs and strategy leaders, AI for Executives & Strategy decisions increasingly depend on understanding how hardware roadmaps and frontier research labs are aligning. The partnership shows that access to compute is no longer a utility; it's a strategic asset that can determine which research agendas scale. For those building or buying AI capabilities, AI Learning Path for CEOs resources can help connect these infrastructure moves to business-level planning. The key takeaway: the companies that own the silicon and the ones that own the safety-critical research are now deeply intertwined, and that will shape what AI products become possible - and profitable - in the next three years.
Your membership also unlocks:
