Safe Superintelligence Inc. (SSI) is the AI research lab that Ilya Sutskever, OpenAI's former chief scientist, cofounded in June 2024 with one stated goal and one product: a safe superintelligence. It sells nothing today. As of November 10, 2026 it has released no model, app, API or research paper, yet it has raised billions of dollars and signed a compute partnership with Nvidia. If you want to know what SSI is, who runs it, how it is funded and what it might release, this guide sets out what the company and credible reporters have actually said.

Everything here was checked on November 10, 2026 against SSI's own site, the Nvidia press release, Sutskever's longest recent interview and wire reporting. Funding figures that come from anonymous sources are labeled as reports. Nothing here is investment advice, and SSI is a private company.

Safe Superintelligence (SSI) at a glance

QuestionShort answer
What it isA private AI research lab focused only on building a safe superintelligence
FoundedJune 19, 2024, by Ilya Sutskever, Daniel Gross and Daniel Levy
Leaders todayIlya Sutskever (CEO) and Daniel Levy (President)
OfficesPalo Alto, California, and Tel Aviv, Israel
ProductsNone released as of November 10, 2026
Money raised$1 billion in 2024 (announced by SSI); about $3 billion by late 2025 (per Sutskever); plus an undisclosed Nvidia investment in July 2026 that Reuters put at $5 billion
Last reported valuation$32 billion (April 2025 round, per Reuters)
ComputeGoogle Cloud TPUs (reported 2025) and Nvidia Vera Rubin systems (announced 2026)

What SSI says it is building

SSI's home page is a single page of plain text. It opens with "Superintelligence is within reach" and calls building safe superintelligence "the most important technical problem of our time." The company describes itself as the first lab aimed straight at that goal, "with one goal and one product: a safe superintelligence." The name, it says, is "our mission, our name, and our entire product roadmap."

Three ideas set it apart from the labs most people know:

  1. No product cycle. SSI says its "singular focus means no distraction by management overhead or product cycles," and that its business model keeps "safety, security, and progress" insulated from commercial pressure. Compare that with OpenAI, which ships ChatGPT, or Anthropic, which ships Claude and lives on subscription and API revenue.
  2. Safety and capability together. The site says SSI treats safety and capabilities "in tandem, as technical problems to be solved," and plans to "advance capabilities as fast as possible while making sure our safety always remains ahead."
  3. A small team. SSI says it is assembling a "lean" team of engineers and researchers. Ctech reported in July 2026 that the company employs "only a few dozen people," and Globes put the count at over 50 in August 2026. SSI has not published a headcount.

"Superintelligence" here means a system far more capable than people across almost every field, not just a better chatbot. Our guides to superintelligence vs AGI and whether superintelligence is possible at all explain the term and the debate around it.

Who runs SSI: Ilya Sutskever and Daniel Levy

Ilya Sutskever cofounded OpenAI and was its chief scientist. In July 2023 he and Jan Leike were named coleads of OpenAI's Superalignment team, set up to work on controlling AI systems much smarter than people. He was also part of the OpenAI board that briefly removed Sam Altman in November 2023. He left OpenAI on May 14, 2024, as The Verge reported, and Leike resigned hours later. OpenAI then dissolved the Superalignment team, according to CNBC.

About a month later, on June 19, 2024, he announced SSI with Daniel Gross, a former Apple AI lead and investor, and Daniel Levy, a former OpenAI researcher (TechCrunch).

Gross did not stay. On July 3, 2025 SSI posted a note from Sutskever on its updates page: Gross was "officially no longer a part of SSI" as of June 29, Sutskever was "now formally CEO," Levy became President, and "the technical team continues to report to me." The note also brushed off "rumors of companies looking to acquire us." CNBC reported that Meta had tried to buy SSI, was turned down and then hired Gross. Meta's AI reorganization from that summer is covered in our explainer on Meta Superintelligence Labs.

Safe Superintelligence SSI funding and valuation

SSI has disclosed only one round itself. The rest comes from reporting, so treat those figures as estimates.

DateWhat happenedSource
Sep 4, 2024Raised $1 billion from NFDG, a16z, Sequoia, DST Global and SV AngelSSI updates page; Reuters reported a $5 billion valuation
Apr 2025$2 billion round led by Greenoaks at a $32 billion valuation; Alphabet and Nvidia joinedReuters, April 2025 (anonymous source)
Apr 2025Google Cloud named SSI a major customer for its TPU chipsReuters, same report
Nov 2025Sutskever said SSI had raised $3 billion in totalDwarkesh Podcast interview
Jul 27, 2026Long term partnership with Nvidia, including an investment of undisclosed size and access to Vera Rubin systemsNvidia press release; Reuters put the investment at $5 billion

The 2024 round is on SSI's own updates page. Reuters reported the $5 billion valuation, which SSI declined to confirm. In April 2025 Reuters reported that Alphabet and Nvidia had invested and that Google Cloud had landed SSI as an anchor customer for its TPUs. Ctech reported the same $2 billion round at $32 billion.

In the Nvidia press release, Nvidia says it "has additionally made an investment in SSI" and that access to Vera Rubin "will allow SSI to increase its compute by an order of magnitude." SSI's updates page says the deal lets it "scale our compute by 10x." Neither company gave an amount. Reuters, citing a person briefed on the deal, reported a $5 billion equity investment.

What is SSI worth? No new valuation has been disclosed since the $32 billion figure from April 2025. Articles that pair the Nvidia money with that number are combining two separate reports, not quoting a new priced round. Total funding is reported anywhere from about $7 billion (Globes) to about $8 billion, depending on how the Nvidia deal is counted.

What SSI has released so far

Nothing a member of the public can use or test. As of November 10, 2026, SSI's site lists three updates in total (the 2024 raise, the 2025 leadership note and the 2026 Nvidia deal) and names no model, paper, demo, waitlist or release date.

What the company says it has is progress behind closed doors. Sutskever told the Dwarkesh Podcast in November 2025, "We've actually made quite good progress over the past year." In the 2026 release he said, "We have research that is worthy of scaling up, and having access to a big NVIDIA computer will let us do so." Nvidia said it invested "after obtaining rare access into the company's closely guarded research." Outsiders cannot check any of that yet.

The August 2026 rumor. Globes reported on August 26, 2026 that investor Gavin Baker had said SSI would launch its first language model within days, and quoted unnamed experts who believed it could update its own weights while reading documents. SSI did not comment, and no launch followed on any official SSI channel through November 10.

How SSI's approach differs from other frontier labs

Sutskever's November 2025 interview is the fullest public account of SSI's thinking. The main points:

  • Research, not scaling. He argued AI has moved from an "age of scaling" back to an "age of research," and called SSI "squarely an 'age of research' company." He said the core open problem is that today's models "generalize dramatically worse than people."
  • A different technical bet. "The main thing that distinguishes SSI is its technical approach," he said, without describing it in detail.
  • The straight shot may bend. Asked whether SSI still plans to go straight to superintelligence, he answered "Maybe," naming two reasons to change course: timelines could be long, and there is value in powerful AI "being out there impacting the world." Even a straight shot, he said, would involve "a gradual release."
  • The goal for alignment. He said the first real superintelligence should be aligned to "care for sentient life."
  • Timeline. Asked when a system that learns as well as a human and then becomes superhuman might arrive, he said "5 to 20" years.
ItemSSIOpenAIAnthropicGoogle DeepMind
Public productsNoneChatGPT, APIClaude, APIGemini models and apps
Published safety rulebookNone publicPreparedness FrameworkResponsible Scaling PolicyFrontier Safety Framework
Stated missionSafe superintelligence onlyEnsure AGI benefits all of humanityAdvanced AI for the long term benefit of humanityBuild AI responsibly to benefit humanity
RevenueNone disclosedSubscriptions and APISubscriptions and APIPart of Alphabet

That table shows the real tradeoff. The bigger labs publish safety frameworks and system cards that outside testers can check, as our comparison of the Preparedness Framework vs Responsible Scaling Policy explains. SSI publishes very little, so its safety claims rest on trust in its people for now. For the other labs, see what Anthropic is, what Google DeepMind does and our explainer on frontier models.

Pros and cons of SSI's model

Pros

  • Fewer commercial distractions: no app to grow, no ads, no quarterly product launches.
  • Led by a researcher with a long track record at the center of modern AI, including OpenAI's GPT work.
  • Explicitly built around safety as an engineering problem, not an afterthought.
  • Large compute access through Google Cloud and Nvidia.

Cons

  • No public output to judge. There is no model, paper or benchmark to test its claims.
  • No published safety framework or outside evaluation, unlike the frameworks at OpenAI, Anthropic and Google DeepMind.
  • Key funding numbers come from anonymous sources, not the company.
  • Its biggest investors are also its chip suppliers, so some of the money likely returns as hardware spending.
  • A cofounder left in 2025, and the lab depends heavily on one leader.

SSI vs "SI" and other meanings of the letters

People searching SSI often find something else. Three quick distinctions:

  • SI in US policy. Since President Trump's September 29, 2026 order, federal agencies say "super intelligence" (SI) where they used to say artificial intelligence. That is a naming rule, not a company. See our explainer on the superintelligence executive order and what SI means.
  • Other "superintelligence" labs. Meta Superintelligence Labs and the Microsoft AI superintelligence team are units inside big companies; see our guide to Microsoft AI Frontiers. SSI is an independent startup.
  • Supplemental Security Income. In the US, SSI also means the Social Security benefit for people with limited income. It has nothing to do with Sutskever's lab.

Who should follow SSI

Follow SSI if you track frontier labs, AI safety research or AI investment, because its first release, whenever it comes, could change how people think about continual learning and alignment. If you just want a capable assistant today, SSI has nothing to offer yet; our ChatGPT vs Claude vs Gemini vs Grok comparison covers the assistants you can actually use. Researchers who want to join the wider field can start with our guide on how to get into alignment and the Alignment Forum.

FAQ

What is Safe Superintelligence (SSI)?

Safe Superintelligence Inc. is a private AI lab founded in June 2024 by Ilya Sutskever, Daniel Gross and Daniel Levy. It says it has one goal and one product, a safe superintelligence, and it has released nothing publicly as of November 10, 2026.

Who owns SSI?

SSI is privately held. Sutskever is CEO and Levy is President. Investors named by SSI or Nvidia include Andreessen Horowitz, Sequoia Capital, DST Global, Greenoaks and Nvidia, and Reuters reported that Alphabet also invested. Ownership percentages have not been published.

How much is SSI worth?

The last reported valuation is $32 billion, from an April 2025 round reported by Reuters. No valuation was disclosed with the July 2026 Nvidia investment, which Reuters put at $5 billion.

Can I use SSI's AI or buy SSI stock?

No to both, for now. SSI has no public model, app or API, and its shares do not trade on any exchange.

Is SSI part of OpenAI?

No. Sutskever cofounded OpenAI but left in May 2024. SSI is a separate company and competes with OpenAI for researchers and compute.

When will SSI release a model?

SSI has not given a date. Sutskever said in November 2025 that even a straight shot plan would involve a gradual release, and a reported August 2026 launch did not happen.