Nvidia invests $5 billion in AI startups
SSI, an AI dark horse that has been dormant for two years, was heavily invested in by Nvidia. On July 27, local time, Nvidia announced that it had agreed to invest approximately US$5 billion in SafeSuperintelligence (hereinafter referred to as "SSI"), an artificial intelligence startup founded by former OpenAI chief scientist Ilya Sutskever. This should have been exciting news, but the market reaction was unexpected: On the day the news was announced, NVIDIA's stock price plummeted nearly 5%, and its market value was overtaken by Apple. Even more puzzling are the investments themselves.
SSI, an AI dark horse that has been dormant for two years, was heavily invested in by Nvidia. On July 27, local time, Nvidia announced that it had agreed to invest approximately US$5 billion in Safe Superintelligence (hereinafter referred to as "SSI"), an artificial intelligence startup founded by former OpenAI chief scientist Ilya Sutskever. This should have been exciting news, but the market reaction was unexpected: On the day the news was announced, NVIDIA's stock price plummeted nearly 5%, and its market value was overtaken by Apple. Even more puzzling are the investments themselves. SSI, a company whose valuation has been pushed up to tens of billions of dollars, has not released a product yet, has never disclosed any research results, and has a team of only a few dozen people. Founder Sutskever also brings his own topics. He was invited by Musk to participate in the founding of OpenAI. In 2023, he was the core figure in the boardroom dispute that removed Sam Altman. He only founded SSI after leaving OpenAI in 2024. Founded by the former founder of OpenAI Without the OpenAI board of directors turmoil in 2023, perhaps SSI today would not have appeared. In December 2015, OpenAI was officially established. Musk, Sam Altman and other Silicon Valley tycoons jointly founded this non-profit organization. Sutskever was invited by Musk to join the founding team. In the nearly ten years since then, Sutskever has been widely regarded as one of the key figures in OpenAI's success with his intuition and judgment on deep learning. The turning point occurred on November 18, 2023. On this day, the OpenAI board of directors suddenly voted to remove CEO Sam Altman without prior notice. Sutskever was one of the core participants. The situation quickly took a turn for the worse: a large number of employees expressed their intention to follow Sam Altman and leave, Microsoft was ready to take over at any time, and the company was on the verge of disintegration. Seeing that the situation was getting out of control, Sutskever reversed his position a few days later over the weekend, publicly apologized for his participation in board actions, and signed a joint letter from employees demanding the return of Sam Altman. In the end, Sam Altman and co-founder Greg Brockman returned to the company, and the storm was finally settled. But for Sutskever, this conflict does not seem to have truly turned the page. On May 15, 2024, he announced on social platforms that he had "decided to leave OpenAI", leaving only a meaningful instruction, hoping that everyone would "create an AGI that is both safe and beneficial." A month later, he founded SSI with Daniel Gross and Daniel Levy. Since its inception, SSI has not used resources to develop commercial products, but has focused on secure superintelligence, focusing on model alignment and general reasoning underlying technologies. After two years of low-key research and development, SSI has officially entered a new stage of large-scale research. To date, SSI has raised approximately US$3 billion in total, with a post-money valuation of US$32 billion. Investors include Andreessen Horowitz, Google parent company Alphabet, Sequoia Capital, Lightspeed Venture Capital and other global first-tier institutions. According to the cooperation agreement, SSI will obtain the right to use NVIDIA's Vera Rubin computing platform, and its computing resources are expected to expand tenfold to support large-scale cutting-edge model training. At the same time, the two companies will establish a joint research and development mechanism and rely on SSI's cutting-edge insights in the fields of general artificial intelligence and model alignment to continue to optimize Nvidia's existing and next-generation AI computing power infrastructure. Valuation: approximately US$32 billion In fact, this is not the first time Nvidia has invested in SSI. It is understood that SSI completed a new round of financing last year at a valuation of approximately US$32 billion. Nvidia was already one of the shareholders at the time, and this time it was an increase in existing relationships. Before this cooperation, SSI's computing power mainly relied on TPU chips from Alphabet's Google. Now the addition of Nvidia's GPU resources also means an adjustment in the supplier structure.
According to people familiar with the matter, the strategic cooperation between SSI and Nvidia took only a few weeks from contact to finalization. According to statements from both parties, SSI will receive a large amount of NVIDIA's flagship GPU resources and will be the first to use NVIDIA's next-generation Vera Rubin computing platform, which will help SSI increase its available computing power to 10 times the current level in the next 12 months. Sutskever said in a statement that the company's research has reached a stage "worthy of scalability," and that it can advance this step after acquiring Nvidia's mainframe computers. Nvidia founder Jensen Huang responded that Sutskever had made groundbreaking contributions to the basic level of modern AI in his early work on AlexNet and other work, and he looked forward to seeing the progress SSI will bring with the help of the new platform. It is worth noting that Nvidia revealed that this investment was made after an in-depth understanding of the research progress of SSI that has been strictly confidential. This is also the closest the outside world has come to get a glimpse of "what exactly SSI is researching" so far. But so far, neither party has said a word. Bind to leading AI companies When SSI received NVIDIA's bet, it was also in the context of intensifying competition in AI infrastructure. Since this year, chip manufacturers such as AMD and Nvidia have tied up with leading AI companies through investment, and the relationship between computing power suppliers and model companies is becoming closer. On July 22, Nvidia’s competitor AMD just announced that it would invest up to US$5 billion in Anthropic; going back further, in March this year, Nvidia had reached a similar large-scale arrangement with Mira Murati’s Thinking Machines Lab. According to reports, the total number of AI infrastructure-related transactions Nvidia is currently negotiating exceeds US$750 billion. In addition to this SSI, it also includes a guarantee of up to US$250 billion for a certain OpenAI data center project and a cooperation with South Korea's SK Group of more than US$500 billion. Some analysts say that in recent years, technology giants including SoftBank and Google are promoting the flow of trillions of dollars in future funds to data center construction and AI research and development. The transaction also highlights the increasingly complex "circular financing" model of the current AI industry - large technology companies provide financial support for infrastructure projects, which in turn will ultimately drive demand for their chips, cloud services and AI products. This makes people worry that this cycle of "chip manufacturers invest in AI companies, and AI companies then turn back to purchase computing power from chip manufacturers" is blurring the boundaries between suppliers and customers. Credit markets were the first to give warning signs. On the day the news was announced, Nvidia's five-year debt default protection cost once rose by about 0.14 percentage points, setting the largest single-day increase since the contract began active trading in November 2025. In the stock market that day, Nvidia closed down 4.99% and AMD closed down 5.17%. NVIDIA's strategy of investing heavily in upstream and downstream operations in recent years has also continued to cause market controversy. Many institutional investors and industry observers have warned that NVIDIA is building a circular financing ecosystem: NVIDIA invests in downstream AI laboratories and computing power operators. After the invested companies obtain funds, they continue to purchase NVIDIA GPUs and computing power services, and the funds flow within the closed loop of the industry chain. This model artificially creates computing power orders, which easily inflates short-term demand and is difficult to reflect the real commercialization capabilities of the end market. Once the implementation of large models fails to meet expectations and computing power rents decline, risk resonance will occur in the entire chain, and NVIDIA's large number of less liquid private equity assets may face large-scale impairment pressure.
According to people familiar with the matter, the strategic cooperation between SSI and Nvidia took only a few weeks from contact to finalization. According to statements from both parties, SSI will receive a large amount of NVIDIA's flagship GPU resources and will be the first to use NVIDIA's next-generation Vera Rubin computing platform, which will help SSI increase its available computing power to 10 times the current level in the next 12 months. Sutskever said in a statement that the company's research has reached a stage "worthy of scalability," and that it can advance this step after acquiring Nvidia's mainframe computers. Nvidia founder Jensen Huang responded that Sutskever had made groundbreaking contributions to the basic level of modern AI in his early work on AlexNet and other work, and he looked forward to seeing the progress SSI will bring with the help of the new platform. It is worth noting that Nvidia revealed that this investment was made after an in-depth understanding of the research progress of SSI that has been strictly confidential. This is also the closest the outside world has come to get a glimpse of "what exactly SSI is researching" so far. But so far, neither party has said a word. Bind to leading AI companies When SSI received NVIDIA's bet, it was also in the context of intensifying competition in AI infrastructure. Since this year, chip manufacturers such as AMD and Nvidia have tied up with leading AI companies through investment, and the relationship between computing power suppliers and model companies is becoming closer. On July 22, Nvidia’s competitor AMD just announced that it would invest up to US$5 billion in Anthropic; going back further, in March this year, Nvidia had reached a similar large-scale arrangement with Mira Murati’s Thinking Machines Lab. According to reports, the total number of AI infrastructure-related transactions Nvidia is currently negotiating exceeds US$750 billion. In addition to this SSI, it also includes a guarantee of up to US$250 billion for a certain OpenAI data center project and a cooperation with South Korea's SK Group of more than US$500 billion. Some analysts say that in recent years, technology giants including SoftBank and Google are promoting the flow of trillions of dollars in future funds to data center construction and AI research and development. The transaction also highlights the increasingly complex "circular financing" model of the current AI industry - large technology companies provide financial support for infrastructure projects, which in turn will ultimately drive demand for their chips, cloud services and AI products. This makes people worry that this cycle of "chip manufacturers invest in AI companies, and AI companies then turn back to purchase computing power from chip manufacturers" is blurring the boundaries between suppliers and customers. Credit markets were the first to give warning signs. On the day the news was announced, Nvidia's five-year debt default protection cost once rose by about 0.14 percentage points, setting the largest single-day increase since the contract began active trading in November 2025. In the stock market that day, Nvidia closed down 4.99% and AMD closed down 5.17%. NVIDIA's strategy of investing heavily in upstream and downstream operations in recent years has also continued to cause market controversy. Many institutional investors and industry observers have warned that NVIDIA is building a circular financing ecosystem: NVIDIA invests in downstream AI laboratories and computing power operators. After the invested companies obtain funds, they continue to purchase NVIDIA GPUs and computing power services, and the funds flow within the closed loop of the industry chain. This model artificially creates computing power orders, which easily inflates short-term demand and is difficult to reflect the real commercialization capabilities of the end market. Once the implementation of large models fails to meet expectations and computing power rents decline, risk resonance will occur in the entire chain, and NVIDIA's large number of less liquid private equity assets may face large-scale impairment pressure.