I've spent the last three years writing about AI power struggles, but the Ilya Sutskever story still doesn't fully make sense to me - and I say that as someone who's tracked this industry daily. A man walks away from one of the most powerful jobs in tech, starts a company with no product, no revenue, and almost no public statements, and investors hand him $32 billion anyway. That's not a typo. That's the story I want to walk you through, fact by verified fact.
1. He Helped Fire Sam Altman - Then Regretted It
Before there was SSI, there was chaos. In November 2023, Ilya Sutskever, then OpenAI's chief scientist, was one of the board members who voted to remove CEO Sam Altman. The decision blew up within days - employees revolted, Altman was reinstated, and Sutskever's role in the boardroom drama became one of the most talked-about moments in Silicon Valley history. According to Business Insider's reporting, Sutskever privately expressed regret over his part in the ouster before formally departing OpenAI in May 2024.
If you want the full blow-by-blow of how deep the rivalry between the two labs runs today, I broke down the entire saga in 15 explosive moments in the Anthropic-OpenAI AI civil war - it's the essential backstory to everything that follows here.
He didn't leave quietly, and he didn't leave to join a rival lab. He left to build something that, on paper, looked like it had no business model at all.
2. One Month Later, He Was Building Something New
Most executives take a "gardening leave" after a departure this dramatic. Sutskever took a month. On June 19, 2024, he announced Safe Superintelligence Inc. (SSI) in a launch covered in detail by TechCrunch - co-founded with Daniel Gross, Apple's former AI lead, and Daniel Levy, a former OpenAI technical staff member.
The company was incorporated in Palo Alto and Tel Aviv, and from day one it was built around a single, almost defiant idea: no distractions, no products, no rat race. Sutskever told Bloomberg's Ashlee Vance at the time that the entire structure was designed to be fully insulated from the pressure of shipping a large, complicated product.
3. "One Goal, One Product": The Strangest Pitch in Tech
Every AI startup pitch deck in 2024 promised a chatbot, an agent, or an API. SSI's pitch, as documented on its official company page, was almost aggressively minimal: "one goal and one product: a safe superintelligence," with team, investors, and business model "all aligned to achieve SSI."
The now-famous closing line from that announcement has been quoted everywhere since: "This way, we can scale in peace."
That single sentence became the entire company culture. No leaderboard chasing. No feature races. No pressure to ship a chatbot before the underlying science is ready.
This isn't the first time a founder has staked everything on a research bet over a product roadmap - I covered a similarly high-stakes wager in Meta's $200 million man: how one researcher became AI's most expensive hire, where the price tag was just as eye-watering.
4. From $5 Billion to $32 Billion in Seven Months
Here's where the story stops sounding like a normal startup and starts sounding like a fever dream. SSI's first funding round, in September 2024, raised $1 billion at a $5 billion valuation - already enormous for a company with no product.
Then, in April 2025, SSI raised another $2 billion in a round led by Greenoaks Capital, pushing its valuation to $32 billion - a jump that Calcalist's CTech reported had "surged sixfold in less than a year." That's for a company that still hadn't shipped anything or published a single research paper.
To put that in perspective: investors weren't betting on a product demo, a beta test, or even a technical paper. They were betting entirely on one man's reputation and a one-line mission statement.
If eye-popping AI valuations detached from real products feel familiar, you'll want to read Google vs. OpenAI: the AI war reshaping the industry - the same "bet on brilliance, not revenue" logic runs through the entire AI arms race right now.
5. Zero Revenue, Zero Product, $32 Billion Anyway
This is the number that stops most people cold. OpenAI is valued near $852 billion and has real, massive revenue. Anthropic sits around $380 billion and is growing fast. SSI is valued at $32 billion with zero revenue. As one analysis from AgentMarketCap put it, the spread implies investors believe frontier AI is a winner-take-most race where fundamental research leadership - not current sales - determines who actually wins.
I dug into whether that kind of confidence is justified, or just hype, in Elon Musk's superintelligence quotes: what he's actually predicting - the same "bet on the breakthrough, not the balance sheet" thinking shows up there too.
In other words, SSI's entire valuation rests on a bet that hasn't paid off yet, might never pay off, and can't be verified by outsiders in any normal way.
6. The Investor List Reads Like an AI Who's-Who
Whatever skepticism you might have about a pre-product startup, it clearly isn't shared by some of the most sophisticated capital in the world. SSI's backers include Andreessen Horowitz, Sequoia Capital, DST Global, and Greenoaks Capital, according to reporting from citybiz - plus strategic stakes from Alphabet and, most recently, Nvidia.
That's not a group of investors chasing a trend. That's a group of investors who each independently decided Sutskever's reputation alone was worth billions of dollars of exposure.
This kind of concentrated, high-stakes capital race across a handful of labs is exactly what I mapped out in The AI arms race: how dangerous is the race to build superintelligence? - SSI is one of the clearest examples of that race in action.
7. Just 50 Employees Split Between Palo Alto and Tel Aviv
Here's the part that genuinely surprised me. Most $32 billion companies have thousands of employees. SSI, according to a detailed breakdown of the company's structure, operates with a lean team of roughly 50 people, split between offices in Palo Alto, California and Tel Aviv, Israel.
There's no massive product org, no customer support team, no sales department. Just a small, tightly vetted research group working on a single multi-year problem with a payoff that, in the company's own words, is "mostly invisible until it isn't."
A small team chasing an outsized, high-stakes goal is a pattern worth watching closely - I explored the same dynamic in Is AI really humanity's last invention? 12 warnings from the scientists who built it.
8. His Own Co-Founder Left for Meta
Even inside SSI's tiny, tight-knit team, the AI talent wars found a way in. Co-founder Daniel Gross, Apple's former AI lead, departed SSI in July 2025 to join Meta's newly formed superintelligence group, as confirmed by Silicon Republic's coverage of the company's later Nvidia deal.
Sutskever stepped up to become CEO in Gross's place, with Daniel Levy taking the title of President. It's a reminder that even the most secretive, mission-driven lab in AI isn't immune to Meta's aggressive, checkbook-first poaching strategy.
Meta's aggressive poaching hasn't been limited to SSI - it's been called out at the highest levels of the industry, as I covered in Big tech CEOs' AI job warnings: what they're really saying.
9. Nvidia Just Handed Him $5 Billion More
Two years of silence ended with a bang. On July 27, 2026, Nvidia announced a $5 billion equity investment in SSI, according to Bloomberg's exclusive reporting - one of the chipmaker's largest single funding commitments of the entire AI boom. Alongside the cash, SSI gained priority access to Nvidia's next-generation Vera Rubin compute platform, expected to boost the lab's available computing power by roughly ten times over the following year.
Nvidia's willingness to bet billions on unproven research labs is a pattern I've tracked closely - see The AI 2027 predictions everyone's talking about for where analysts think this compute race is actually heading.
Jensen Huang didn't hide his reasoning either, crediting Sutskever's foundational work on AlexNet as the reason Nvidia wanted in.
10. "A Different Mountain to Climb"
So what is SSI actually doing with all that money and compute? According to a detailed financial breakdown of the company, Sutskever has told people close to him that SSI is pursuing a direction "distinct from the methods used at OpenAI" - something he's described as "a different mountain to climb."
That's a notable admission from the person who helped build the scaling playbook - bigger models, bigger datasets, bigger compute - that defined GPT-3 and GPT-4. He's now betting that playbook has hit a ceiling, and that the next real breakthrough needs a genuinely different approach, not just a bigger version of the same one.
11. Sutskever's Three Eras of AI
Sutskever has reportedly framed the last 15 years of AI progress into three distinct phases, as detailed by The Globe and Mail's coverage of SSI's trajectory: a research era from roughly 2012 to 2020, a scaling era from 2020 to 2025, and a return to a research-driven era starting around 2026 - where new ideas, not more GPUs, drive the next real gains.
This shift away from "just add more compute" thinking connects directly to concerns I covered in Superintelligence extinction risk: what scientists are actually warning about - the idea that raw scale alone might not be a safe path forward.
If Sutskever is right, every lab still racing to build bigger and bigger models might be climbing the wrong mountain entirely.
12. Two Years of Total Silence Was the Strategy
For two full years, SSI published almost nothing. No benchmarks. No papers. No product roadmap. No leaderboard appearances. In an industry obsessed with weekly announcements, that kind of silence is almost unheard of - and according to Enterprise DNA's analysis of the Nvidia deal, that silence was "entirely intentional."
The reasoning: SSI doesn't want to be held to capability milestones set by outside expectations. It wants to actually solve the problem first - a philosophy that stands in sharp contrast to nearly every other major AI lab in 2026.
13. The "Bunker" Comment That Still Haunts Him
If you want to understand how seriously Sutskever takes the danger of the technology he's building, this is the fact that says it best. According to excerpts from Karen Hao's book "Empire of AI," published by Fortune, Sutskever told OpenAI researchers in a 2023 meeting: "We're definitely going to build a bunker before we release AGI."
When a startled colleague asked him to clarify, he reportedly confirmed it matter-of-factly, adding that getting into the bunker would be optional. Other sources told Hao that some researchers close to Sutskever believed achieving AGI would trigger a "rapture" - his word, not theirs.
This isn't the only unsettling real-world example of AI systems and their builders behaving in ways nobody fully expected - I documented several verified cases in AI systems that refused shutdown: real examples from 2026.
Whatever you think of the bunker comment, it tells you something important: the man now building "safe" superintelligence has never pretended the underlying technology is anything less than potentially catastrophic.
14. He Built the Paper That Started the Deep Learning Era
Sutskever's credibility isn't just reputation - it's documented history. In 2012, as a graduate student, he co-authored the AlexNet paper, the breakthrough that is widely credited with kickstarting the modern deep learning revolution. He went on to lead the core technical teams behind GPT-2, GPT-3, and GPT-4 at OpenAI.
Nvidia CEO Jensen Huang referenced that legacy directly when announcing the company's $5 billion investment, stating in Nvidia's official partnership announcement that Sutskever "has pioneered fundamental breakthroughs at the foundation of modern AI, beginning with AlexNet."
That track record is the entire reason SSI's valuation makes any sense at all. Investors aren't buying a product. They're buying a 14-year history of being right about where AI was headed next.
15. Why SSI's Valuation Might Be the Real AI Warning Sign
Step back from the individual facts, and a bigger pattern emerges. SSI's valuation has stayed locked at $32 billion since April 2025, even as total funding climbed to $8 billion, according to a detailed valuation breakdown from ValueAddVC. Zero revenue. No shipped model. No published research. And yet some of the smartest capital on earth keeps writing checks.
Not everyone agrees this level of confidence in superintelligence timelines is justified - I laid out the strongest counterarguments in Superintelligence risk, debunked: what the skeptics are saying, which is worth reading alongside this piece for balance.
Whether SSI eventually ships a model that justifies the bet, or turns out to be the most expensive research project in history with nothing to show for it, one thing is already clear: the amount of money chasing "safe" superintelligence right now says a lot about how seriously the industry's biggest players take the risk of the "unsafe" version arriving first.
Frequently Asked Questions (FAQ)
What is Ilya Sutskever's Safe Superintelligence (SSI)?
Safe Superintelligence Inc. is an AI research lab founded by Ilya Sutskever in June 2024, focused solely on building a safe superintelligent AI system, with no other commercial products.
How much is Safe Superintelligence worth?
SSI was valued at $32 billion as of an April 2025 funding round, a valuation that has held steady even as total funding rose to roughly $8 billion by mid-2026.
Why did Ilya Sutskever leave OpenAI?
Sutskever left OpenAI in May 2024, one year after being involved in the unsuccessful November 2023 attempt to remove CEO Sam Altman, a decision he later reportedly regretted.
Does SSI have a product yet?
As of the most recent reporting, SSI has not released a commercial product, published significant research, or announced a launch date - its entire public identity remains centered on its research mission.
Who are Ilya Sutskever's co-founders at SSI?
SSI was co-founded by Sutskever, Daniel Gross (former Apple AI lead), and Daniel Levy (a former OpenAI technical staff member). Gross left for Meta's superintelligence group in July 2025.