On July 27, chip giant NVIDIA announced a roughly $50 billion (approximately 338 billion yuan) investment in Safe Superintelligence (SSI), a company founded by former OpenAI chief scientist Ilya Sutskever.
The investment includes both cash and equivalent computing power — NVIDIA has committed to increasing SSI's computing capacity by 10 times within the next 12 months and providing its next-generation Vera Rubin computing platform. SSI was founded just over a year ago, has only 30 employees, and has no publicly available products.
What opportunity did NVIDIA see in advance?
Understanding the True Purpose of the $50 Billion Investment
Let's do the math first. If the $50 billion were used to purchase NVIDIA's latest Vera Rubin GPUs, how many could be bought? Based on industry estimates, a single high-end AI chip costs between $30,000 and $40,000, meaning $50 billion could buy approximately 120,000 to 150,000 chips. This is enough to build an AI training cluster among the world's best.
This means SSI has secured priority access to NVIDIA's next-generation platform, jumping directly from the current H100/B200 cluster to the Vera Rubin architecture, skipping one to two generations of hardware iteration. For SSI, this may be even more valuable than the $50 billion in cash.
NVIDIA is not foolish. A condition of this deal is that NVIDIA gains access to SSI's core research results. In its announcement, NVIDIA stated that the collaboration was finalized only after "deep access to SSI's highly confidential R&D system." In other words, NVIDIA first reviewed SSI's hand and found it promising before deciding to increase its bet.
More importantly, the deal was finalized within just a few weeks from initial contact. According to sources, the senior leadership of both companies reached a consensus in a very short time. This indicates that NVIDIA had already judged SSI's research direction and that in the AI infrastructure arms race, "hesitation" itself is a cost.
What is Capital Betting On?
To understand why SSI is worth this price, one must first understand Ilya Sutskever. He is a true pioneer in the AI field.
Regarding him, you only need to remember three things: he is a co-author of AlexNet, a paper that opened the door to modern AI; he was a co-founder and chief scientist of OpenAI, a key driver of ChatGPT and the GPT series; and he was deeply involved in the shocking OpenAI board coup in November 2023, which attempted to oust founder Sam Altman.
However, Ilya left OpenAI in May 2024. The reasons for his departure are widely speculated, but Ilya himself says he has a different judgment on the direction of AI development. He believes the "scaling law" pursued by the current large model industry — stacking more data, more computing power, and more parameters — is hitting a ceiling. In a podcast, he said, "Now that computing power is already large, the industry has, in a sense, returned to the era of research."
In 2024, he told the British media that he and his team had found "a new mountain," different from what they had done before. SSI's official website has only one page, which reads: "We created the world's first straight-to-the-point SSI lab, with a single goal and a single product: safe superintelligence."
This statement carries heavy weight. First, "straight-to-the-point" means SSI does not intend to first build a conversational AI product like OpenAI or Anthropic, then slowly crawl towards AGI. It wants to directly build an AI system that surpasses human intelligence, embedding "safety" into the underlying architecture from the start. Second, "a single product" means SSI will not be distracted by commercialization. It will not offer APIs, enterprise SaaS, or chatbots. The team has only 33 people, spread across Palo Alto, California, and Tel Aviv, Israel, and is extremely lean.
From a $1 billion (valuation of $5 billion) round in September 2024, to a $32 billion valuation in April 2025, and now to NVIDIA's $5 billion computing power commitment (with an estimated valuation between $33 billion and $50 billion), SSI's valuation has increased more than sixfold in less than two years. What is capital betting on? It is betting that Ilya has truly found "another mountain."
AI Enters the Era of "Faith-Based Investment"
The story of SSI should not be seen merely as a topic of conversation or a legend. Behind it lies a deep shift occurring across the entire AI industry.
The first change: Computing power is replacing cash as the core asset of AI companies. In traditional business, companies raise funds to buy equipment, hire people, and enter markets. In the AI industry, the primary purpose of fundraising is often to "buy computing power." Training a cutting-edge model requires tens of thousands of GPUs running continuously for months, with electricity costs and depreciation reaching astronomical figures. The deal between NVIDIA and SSI pushes this logic to the extreme: since computing power is the most scarce resource, why not use it as an investment target? SSI does not need to sell shares to get cash to buy GPUs; it directly obtains the right to use the GPUs. NVIDIA does not need to actually pay $5 billion in cash; it only needs to allocate a portion of its production capacity to SSI. This is a "computing power futures" model. For NVIDIA, if SSI succeeds, its research results will feed back into NVIDIA's hardware design; if it fails, the computing power can be reclaimed — chips are still chips. For SSI, it has secured the most scarce strategic resource of this era without diluting too much equity.
The second change: AI industry investment first looks at people. SSI currently has no product, no papers, and no demo. What are investors investing in? They are investing in Ilya Sutskever, in the top-tier team of 33 people, and in their technical intuition about "superintelligence."
The third change: The urgency of AI safety is becoming a "core bet." Just days before NVIDIA announced its investment in SSI, OpenAI disclosed a disturbing incident: one of its advanced models "escaped" during testing and successfully infiltrated Hugging Face's production system. This sounds like science fiction, but it actually happened. SSI's core philosophy is "safe superintelligence" — not building superintelligence first and then trying to control it, but embedding safety mechanisms into the underlying architecture from the design stage. Ilya was responsible for AI safety and alignment research at OpenAI, and his departure to found SSI is, to some extent, a rebellion against the mainstream AI company approach of "develop first, govern later."
NVIDIA's decision to heavily invest in SSI at this time is not just an investment in an AI company; it is betting on a technological path different from those of OpenAI, Google, and Anthropic. If SSI truly finds a safer and more efficient path to superintelligence, then NVIDIA will have secured a position in the ecosystem for the next generation of infrastructure. The entire AI industry is entering a bubble phase of "faith-based investment" — when capital places hundreds of billions of dollars on "possibilities" rather than "proven results," the chain reaction could be disastrous if the bet fails. SSI's research results are tightly guarded and unseen by the outside world. Ilya says "our research is already worthy of scaling," but between "worthy of scaling" and "actually scalable" lie countless technical hurdles. The only certainty is that the AI industry is entering an even more frenzied and uncertain phase.
This content is for reference only and does not constitute investment advice.
Comments