Skip to content
← All organizations

Frontier AI research lab · founded Jun 2024 · Palo Alto, USA, and Tel Aviv, Israel

Safe Superintelligence Inc.

Safe Superintelligence Inc. is the lab Ilya Sutskever started a month after leaving OpenAI, on the premise that one company with "one goal and one product" could build superintelligence while keeping safety ahead of capability and staying clear of product cycles. It is a small private American company with offices in Palo Alto and Tel Aviv. It has raised billions on its founders' reputations, runs on Google TPUs and now Nvidia hardware, and has published no model or paper. It is the test case for the argument that a lab can build in private and release only when ready.

History

A straight shot

Ilya Sutskever announced on 14 May 2024 that he was leaving OpenAI, six months after voting with its board to remove Sam Altman, for "a project that is very personally meaningful to me." On 19 June he, Daniel Gross and Daniel Levy named it. Gross had sold his search startup Cue to Apple and run AI projects there before becoming a Y Combinator partner and an investor; Levy had led OpenAI's optimization team. Gross became chief executive.

The founding statement, still the company's homepage, was a single short page. "Building safe superintelligence (SSI) is the most important technical problem of our time," it began. "We have started the world's first straight-shot SSI lab, with one goal and one product: a safe superintelligence." It said the company would "advance capabilities as fast as possible while making sure our safety always remains ahead," so that "we can scale in peace," and that its business model would keep "safety, security, and progress" insulated "from short-term commercial pressures." It described SSI as "an American company with offices in Palo Alto and Tel Aviv" and asked for "a lean, cracked team" of engineers and researchers.

Sutskever told Bloomberg that the safety he meant was engineering safety: "By safe, we mean safe like nuclear safety as opposed to safe as in 'trust and safety.'" The announcement did not say how SSI would earn money or what technical approach it would take, and the company has kept both largely private since.

Money before a product

In September 2024 the ten-person company announced 1 billion dollars from NFDG (Gross's fund with Nat Friedman), Andreessen Horowitz, Sequoia, DST Global and SV Angel, at a valuation that people close to the deal put at 5 billion dollars. Gross said investors had to support "a straight shot to safe superintelligence and in particular to spend a couple of years doing R&D on our product before bringing it to market." Sutskever explained the new venture as "a mountain that's a bit different from what I was working on," and questioned the scaling program he had championed at OpenAI: "Everyone just says scaling hypothesis. Everyone neglects to ask, what are we scaling?" Three months later he told the NeurIPS conference that "pre-training as we know it will unquestionably end" because the supply of internet text was running out.

Investors kept coming. In February 2025 TechCrunch reported that Greenoaks was leading a round at a 30 billion dollar valuation, and in April reports put the finished round at 2 billion dollars and a 32 billion dollar valuation, with Alphabet and Nvidia also backing the company; SSI did not comment. On 9 April 2025 Google Cloud announced that SSI was using its TPU chips for research, and a source told TechCrunch that Google Cloud was SSI's primary computing provider.

Meta's offer and a new CEO

In June 2025, during Mark Zuckerberg's hiring drive for what became Meta Superintelligence Labs, CNBC reported, citing unnamed sources, that Meta had recently tried to acquire SSI, that Sutskever had rebuffed both the offer and an attempt to hire him, and that Zuckerberg had then negotiated with Gross instead. Meta hired Gross and Friedman and took a stake in their fund.

Gross's last day was 29 June. On 3 July Sutskever wrote to staff and investors: "I am now formally CEO of SSI, and Daniel Levy is President. The technical team continues to report to me." He added, "You might have heard rumors of companies looking to acquire us. We are flattered by their attention but are focused on seeing our work through," and "We have the compute, we have the team, and we know what to do." Gross wrote that the company's "future is very bright, and I expect miracles to follow." In January 2026 Gross became co-leader of Meta Compute, Meta's data center buildout.

The two accounts of the split differ in emphasis. Asked in November 2025 whether a co-founder's exit meant SSI lacked breakthroughs, Sutskever said the company had been raising money at a 32 billion dollar valuation when "Meta came in and offered to acquire us, and I said no. But my former cofounder in some sense said yes." He said Gross had gained "a lot of near-term liquidity" and was "the only person from SSI to join Meta."

An age-of-research company

Sutskever's interview with Dwarkesh Patel on 25 November 2025 is the fullest account of SSI's plans. He said SSI had raised 3 billion dollars and that this bought more research compute than it seemed, because rivals spend much of theirs on serving products; that "we are squarely an 'age of research' company"; that its distinguishing feature was "a different technical approach"; and that the question of how SSI would make money would "reveal itself." Asked whether SSI still planned to go straight to superintelligence, he answered, "Maybe," naming two reasons to change course: timelines might prove long, and there was value in powerful AI "being out there impacting the world." Even in the straight-shot case, he said, "Gradualism would be an inherent component of any plan."

On 27 July 2026 Nvidia and SSI announced a long-term partnership that gives SSI access to Nvidia's Vera Rubin platform and, in SSI's words, lets it "scale our compute by 10x." Nvidia said it had entered the deal "after obtaining rare access into the company's closely guarded research," and Bloomberg reported Nvidia's investment at 5 billion dollars. "We have research that is worthy of scaling up," Sutskever said. In early August investor Gavin Baker said on a podcast that "SSI says that they'll come out with their model in August." SSI never confirmed that, and as of late September 2026 its updates page listed no release after the Nvidia partnership.

Timeline

  1. May 2024

    Sutskever leaves OpenAI

    OpenAI's co-founder and chief scientist departs for a project he calls personally meaningful. Source

  2. Jun 2024

    Safe Superintelligence Inc. announced

    Sutskever, Daniel Gross and Daniel Levy describe a "straight-shot" lab with one goal and one product. Source

  3. Sep 2024

    First 1 billion dollars raised

    Andreessen Horowitz, Sequoia, DST Global, SV Angel and NFDG invest at a reported 5 billion dollar valuation. Source

  4. Dec 2024

    Sutskever tells NeurIPS pre-training will end

    In his Test of Time talk he says the field has reached "peak data" and must find new approaches. Source

  5. Feb 2025

    Reported round at 30 billion dollars

    TechCrunch reports Greenoaks is leading a raise of more than 1 billion dollars, six times the prior valuation. Source

  6. Apr 2025

    Google Cloud TPUs

    Google Cloud says SSI is using its TPU chips for research; a source calls Google SSI's primary compute provider. Source

  7. Apr 2025

    Reported valuation of 32 billion dollars

    SSI is reported to have raised 2 billion dollars more in a round led by Greenoaks; the company does not comment. Source

  8. Jun 2025

    Meta's acquisition attempt reported

    CNBC reports that Sutskever rebuffed an offer from Meta, which then moved to hire Gross. Source

  9. Jun 2025

    Daniel Gross leaves

    The co-founder and chief executive departs to join Meta Superintelligence Labs. Source

  10. Jul 2025

    Sutskever becomes CEO

    Daniel Levy becomes president, and Sutskever tells investors SSI is not for sale. Source

  11. Nov 2025

    Sutskever calls SSI an "age of research" company

    In a long interview he says SSI has raised 3 billion dollars and has a different technical approach. Source

  12. Jul 2026

    Nvidia partnership

    Access to Nvidia's Vera Rubin platform and an Nvidia investment reported at 5 billion dollars. Source

  13. Aug 2026

    Investor claims an August model release

    Gavin Baker says on a podcast that SSI plans a model that month; SSI does not confirm, and no release follows by late September. Source

Where it stands

The lab's own stated positions, from its charters, policies and official statements.

What are the biggest risks from AI?

SSI's June 2024 founding statement treats safety and capability as engineering problems to be solved together, with safety kept ahead of capabilities, and says its structure shields that work from commercial pressure.

“We approach safety and capabilities in tandem, as technical problems to be solved through revolutionary engineering and scientific breakthroughs.”

Safe Superintelligence Inc. founding statement, 2024 · Compare the people on this

Can language models reach general intelligence?

Speaking as CEO in November 2025, Sutskever said current methods, scaled-up pre-training and reinforcement learning on language models, will keep improving but stall short of human-like learning, and that SSI is betting on new research into generalization.

“I think what people are doing right now will go some distance and then peter out.”

Dwarkesh Podcast, "Ilya Sutskever, We're moving from the age of scaling to the age of research," November 2025, 2025 · Compare the people on this

How close is human-level AI?

Asked in November 2025 how long until a system that learns as well as a human and so becomes superhuman, Sutskever gave a range of five to twenty years.

“I think like 5 to 20.”

Dwarkesh Podcast, November 2025, 2025 · Compare the people on this

Who should set the rules for AI?

Sutskever has said that showing the public increasingly powerful AI is what will prompt governments to act and rival labs to cooperate on safety, and that this view may change SSI's plan to release nothing before superintelligence.

“as AI continues to become more powerful, more visibly powerful, there will also be a desire from governments and the public to do something.”

Dwarkesh Podcast, November 2025, 2025 · Compare the people on this

People

Here now

  • Ilya Sutskever

    Co-founder; CEO since July 2025, 2024-present

    Co-author of AlexNet and sequence-to-sequence learning, OpenAI co-founder and chief scientist until 2024, a board member who voted to fir...

  • Daniel Levy Brief profile

    Co-founder; president since July 2025, 2024-present

    Led OpenAI's optimization team and co-led vision for GPT-4, then co-founded Safe Superintelligence as its president.

Formerly here

  • Daniel Gross Brief profile

    Co-founder and CEO, 2024-2025

    Co-founded Safe Superintelligence with Ilya Sutskever and was its first CEO, then left for Meta in June 2025.

Sources

Last verified 2026-09-25.

  1. Safe Superintelligence Inc. founding statement (June 2024, checked September 2026)
  2. Safe Superintelligence Inc. updates (September 2024 funding, July 2025 CEO message, July 2026 Nvidia partnership)
  3. Ilya Sutskever's Safe Superintelligence Inc. and NVIDIA announce long-term strategic partnership (NVIDIA, July 2026)
  4. Ilya Sutskever, We're moving from the age of scaling to the age of research (Dwarkesh Podcast, November 2025)
  5. Ilya Sutskever, OpenAI co-founder and longtime chief scientist, departs (TechCrunch, May 2024)
  6. He tried to oust OpenAI's CEO. Now, he's starting a 'safe' rival (CNN, June 2024)
  7. OpenAI co-founder Ilya Sutskever's new venture secures 1 billion dollars (Business Today, September 2024)
  8. OpenAI co-founder Ilya Sutskever says the way AI is built is about to change (The Verge, December 2024)
  9. Safe Superintelligence is reportedly close to raising roughly 1 billion dollars (TechCrunch, February 2025)
  10. Ilya Sutskever taps Google Cloud to power his AI startup's research (TechCrunch, April 2025)
  11. Safe Superintelligence reportedly valued at 32 billion dollars (TechCrunch, April 2025)
  12. Meta tried to buy Safe Superintelligence, hired CEO Daniel Gross (CNBC, June 2025)
  13. Ilya Sutskever becomes CEO of Safe Superintelligence after Meta hired Gross (CNBC, July 2025)
  14. "I expect miracles": Gross hails Safe Superintelligence's future as he joins Meta (Ctech, July 2025)
  15. Ilya Sutskever's Safe Superintelligence partners with Nvidia to scale its AI research (TechCrunch, July 2026)
  16. SSI to come out with their model in August, claims Gavin Baker (OfficeChai, August 2026)