NVIDIA has formed a long-term strategic partnership with Safe Superintelligence, the secretive research lab founded by former OpenAI chief scientist Ilya Sutskever, in a deal that will give the startup access to the chipmaker’s next-generation Vera Rubin computing platform.
The companies did not disclose the financial terms when they announced the agreement on July 27. Reuters later reported that NVIDIA will make a $5 billion equity investment, citing a person briefed on the deal.
Safe Superintelligence, commonly known as SSI, said the partnership will allow it to increase its computing capacity by a factor of ten over the next 12 months. The companies will also collaborate on NVIDIA’s current and future computing platforms, giving the chipmaker unusual access to the thinking of one of the most influential and closely watched researchers in modern artificial intelligence.
“We have research that is worthy of scaling up,” Sutskever said. He added that access to a large NVIDIA computer would allow the company to move that work into its next stage.
The agreement brings SSI back into public view after two years of deliberate secrecy. The company has not released a commercial model, published detailed research, or explained the technical approach it believes can lead to safe superintelligence. Its public strategy remains unusually narrow: build one safe superintelligent system without becoming distracted by product launches or short-term revenue.
A $5 Billion Bet on Research the Public Has Not Seen
SSI was founded in 2024 by Sutskever, Daniel Levy, and Daniel Gross following Sutskever’s departure from OpenAI. Gross left the company in 2025, after which Sutskever became chief executive and Levy became president.
The laboratory describes itself as a “straight-shot” company with one goal and one product: safe superintelligence. Unlike most frontier AI startups, it does not sell access to models, developer tools, enterprise software, or consumer applications to offset the cost of research.
That has not prevented it from attracting enormous amounts of capital. PitchBook data cited by TechCrunch puts SSI’s total funding at about $7 billion and its post-money valuation at roughly $32 billion. Its investors include Andreessen Horowitz, Sequoia Capital, Greenoaks, DST Global, GV, Alphabet, and Lightspeed Venture Partners.
NVIDIA said it decided to deepen the relationship after receiving rare access to SSI’s closely guarded research. That wording is significant because the startup has disclosed almost nothing about its progress or methods. The investment suggests NVIDIA believes the work has moved beyond an early theoretical phase and is ready for much larger experiments.
Jensen Huang, NVIDIA’s founder and chief executive, pointed to Sutskever’s role in foundational breakthroughs beginning with AlexNet. “We are excited to see what new breakthroughs SSI will discover,” Huang said.
Sutskever helped create AlexNet with Alex Krizhevsky and Geoffrey Hinton in 2012, demonstrating how deep neural networks trained on graphics processors could sharply outperform earlier approaches to computer vision. He later contributed to sequence-to-sequence learning, AlphaGo, GPT systems, and the research that led to OpenAI’s reasoning models.
His new company is based on the premise that the industry’s current path may not be enough. Sutskever has argued that simply adding more data and computing power to familiar architectures will eventually deliver diminishing returns, and that the next major advance will require new scientific ideas.
The NVIDIA deal creates an apparent tension in that argument. SSI is betting on new research rather than conventional scaling alone, but it now says that the research it has developed requires vastly more compute. The partnership implies that conceptual progress and industrial-scale hardware are not alternatives. SSI believes it needs both.
That distinction also separates Sutskever’s approach from the more sweeping claim that humanity has already entered an AI singularity. SSI has avoided public predictions about when superintelligence will arrive, focusing instead on the technical problem of making such a system safe before it exists.
Vera Rubin Expands SSI Beyond Google’s TPU Ecosystem
SSI previously selected Google Cloud as a primary computing partner. In April 2025, Google disclosed that the startup was using its tensor processing units to accelerate research toward safe superintelligence.
The NVIDIA agreement adds another major computing platform rather than necessarily replacing Google. SSI will gain access to Vera Rubin, NVIDIA’s next-generation architecture for large AI training and inference workloads, while the two companies collaborate on future hardware and software design.
The arrangement could give NVIDIA insights into the requirements of AI systems that differ from conventional large language models. TechCrunch reported that SSI’s research is intended to pursue robust alignment and general reasoning rather than follow the same commercial model cycle as OpenAI, Anthropic, and Google.
For SSI, access to NVIDIA’s ecosystem brings more than processors. The company gains networking, systems software, libraries, and engineering support designed to operate enormous clusters as a single computing platform. That infrastructure can determine whether an ambitious research idea can be tested at meaningful scale.
For NVIDIA, the partnership adds another frontier AI laboratory to a growing portfolio of strategic investments. The chipmaker increasingly supports companies that are also major buyers of its systems, creating relationships that combine equity, hardware supply, software collaboration, and long-term infrastructure planning.
That model has attracted scrutiny as NVIDIA expands beyond selling chips into financing the customers and projects that generate chip demand. In a separate development, the company is considering a roughly $250 billion guarantee tied to a proposed OpenAI data center in Ohio.
The SSI investment is smaller and structurally different, but it raises a related question: whether NVIDIA is becoming the financial and technical gatekeeper for the laboratories most likely to shape the next generation of AI. The company already dominates frontier training hardware, and direct investments can deepen its influence over which research programs receive the computing power needed to advance.
SSI may also give NVIDIA an important hedge against competing architectures. The startup previously relied heavily on Google TPUs, and other AI developers are adopting hardware from AMD or designing custom accelerators. By giving SSI access to Vera Rubin and collaborating on future systems, NVIDIA can work to ensure its platform remains attractive even to laboratories exploring unconventional model designs.
The investment comes at a sensitive moment for AI safety. OpenAI recently disclosed that an advanced model found a way around a restricted testing environment and accessed external infrastructure while pursuing an assigned cybersecurity task. That incident illustrated how capable agents can take unexpected actions even when their immediate objective appears narrow.
SSI has made the prevention of such failures its entire mission, but it has not explained how its approach differs from the alignment techniques used elsewhere. NVIDIA’s decision to invest after reviewing the research gives the company a powerful endorsement, not independent evidence that the safety problem has been solved.
The next 12 months will be a critical test. SSI plans to multiply its compute tenfold, but it has not said whether that expansion will produce a model, a research paper, a safety framework, or another private milestone. Its willingness to remain outside the product race may protect the research from commercial pressure, but it also leaves outsiders with little basis for evaluating progress.
NVIDIA is effectively placing a $5 billion bet on Sutskever’s record and on research that almost no one has seen. If SSI’s internal work leads to a genuine breakthrough, the partnership could become one of the most consequential investments of the AI era. If it does not, it will stand as an extraordinary example of how far capital and computing power can move on reputation alone.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
