Safe Superintelligence Inc. (SSI): Ilya Sutskever’s $32B Bet With No Product

Safe Superintelligence Inc. explained: who founded SSI, how much it has raised, its $32B valuation, why it ships no product, and what Sutskever has said about its plan.

By Izhar Ahmad Danish · · 8 min read

Safe Superintelligence Inc. (SSI): Ilya Sutskever’s $32B Bet With No Product: Time of Superintelligence cover image
On this page
  1. Why Sutskever left OpenAI to start SSI
  2. Who founded Safe Superintelligence?
  3. What is SSI trying to build?
  4. Funding and valuation timeline
  5. Why Daniel Gross left for Meta
  6. What has Sutskever said about SSI’s approach?
  7. SSI vs other superintelligence labs
  8. What does “safe” mean at SSI?
  9. Can a lab with no product win the race?
  10. Common confusions about “SSI”
  11. Our take: what to watch
  12. Frequently asked questions
  13. Sources

Safe Superintelligence Inc. (SSI) is the AI lab Ilya Sutskever founded after leaving OpenAI in 2024. It has one goal, building a superintelligence that is safe, and it has promised to release nothing until it gets there. With no product and no revenue, investors still valued it at about $32 billion in 2025.

Short answer: Safe Superintelligence is a private research company in Palo Alto and Tel Aviv, founded on June 19, 2024 by Ilya Sutskever, Daniel Gross and Daniel Levy. It has raised billions of dollars to pursue a “straight shot” to superintelligence: no chatbot, no API, no products along the way. As of October 2026 it has not shown a public model.

Why Sutskever left OpenAI to start SSI

Short answer: SSI grew out of Sutskever’s years leading safety work at OpenAI and the November 2023 board crisis, after which his position there became untenable.

Three episodes explain why SSI exists:

  1. The Superalignment team (July 2023). OpenAI announced a new team, co-led by Sutskever and Jan Leike, to solve the problem of controlling AI systems “much smarter than us” within four years. OpenAI pledged 20% of its compute to the effort. It was the clearest sign that OpenAI’s chief scientist considered superintelligence a near-term engineering problem.
  2. The board crisis (November 2023). Sutskever was one of the OpenAI board members who voted to fire CEO Sam Altman. Within days, almost all staff threatened to quit, Sutskever said he deeply regretted his part, and Altman returned. Sutskever left the board and was rarely seen at the company afterwards.
  3. The exit (May 2024). Sutskever announced his departure on May 14, 2024. Leike quit the same week, writing that at OpenAI “safety culture and processes have taken a backseat to shiny products.” The Superalignment team was dissolved soon after.

SSI is, in effect, Sutskever’s answer to the third point: a lab built so that products can’t take priority, because there are no products.

Who founded Safe Superintelligence?

Three people founded SSI. One of them left about a year later.

Founder Background Role at SSI
Ilya Sutskever Co-founder and chief scientist of OpenAI; co-author of AlexNet (2012) with Geoffrey Hinton Co-founder; CEO since July 2025
Daniel Gross Former Apple AI lead, Y Combinator partner, investor Co-founder and first CEO; left for Meta in July 2025
Daniel Levy Former OpenAI researcher Co-founder; president

Sutskever is the reason SSI matters. He was one of the most respected researchers in deep learning, a student of “godfather of AI” Geoffrey Hinton, and the OpenAI board member who voted to remove Sam Altman in November 2023 before reversing himself. He left OpenAI in May 2024, and SSI was announced a month later.

What is SSI trying to build?

Short answer: a single system, a safe superintelligence, built as a pure research effort with no commercial products until it is done.

The founding announcement on ssi.inc set out the plan in a few lines:

  • “We have started the world’s first straight-shot SSI lab, with one goal and one product: a safe superintelligence.”
  • Safety and capabilities are treated “as technical problems to be solved through revolutionary engineering and scientific breakthroughs.”
  • The business model means safety and progress are “insulated from short-term commercial pressures,” with no “management overhead or product cycles.”

That is a deliberate contrast with OpenAI, which Sutskever helped build and which now funds its research with ChatGPT subscriptions and enterprise deals. SSI’s argument is that shipping products creates pressure to cut corners. Critics answer that a lab with no users also gets no real-world feedback about how its systems behave.

Funding and valuation timeline

SSI has no revenue, so its value rests entirely on investors’ belief in Sutskever and his team.

Date Event Valuation (reported)
Jun 19, 2024 SSI announced; offices in Palo Alto and Tel Aviv —
Sep 2024 Raises $1 billion from a16z, Sequoia, DST Global, SV Angel and NFDG (Gross and Nat Friedman’s fund) ~$5 billion
Mar–Apr 2025 New round led by Greenoaks Capital, about $2 billion; Alphabet and Nvidia reported as investors ~$30–32 billion
Apr 2025 Google Cloud says SSI will use its TPU chips for research —
Jul 2025 Daniel Gross leaves to join Meta’s superintelligence effort; Sutskever becomes CEO —
Oct 2026 No public model, product or revenue reported Latest confirmed: ~$32 billion

A sixfold jump in valuation in about six months, with no product, has few precedents. It shows how much investors were willing to pay in 2025 for what they saw as a credible path to artificial super intelligence.

Unconfirmed: some 2026 blogs claim new investments and a planned model release in 2026. We have not found confirmation from SSI or a major news outlet, so this page does not count them.

Why Daniel Gross left for Meta

In mid-2025 Mark Zuckerberg was building Meta Superintelligence Labs and hiring aggressively. Reports said Meta first tried to acquire SSI and was turned down. Meta then hired Daniel Gross, and his investment partner Nat Friedman, instead.

Sutskever confirmed the change in a July 2025 post: Gross’s time at SSI had come to an end, Sutskever would become CEO and Levy president, and the company would keep building. “We have the compute, we have the team, and we know what to do,” he wrote. He also said SSI was not looking to be acquired.

What has Sutskever said about SSI’s approach?

Short answer: he thinks simply scaling today’s models is no longer enough, and that new research ideas, especially about how models generalize, are the bottleneck.

Sutskever has rarely given interviews since founding SSI. The main points he has made publicly:

  • “The age of scaling” is ending. In a widely shared November 2025 podcast interview with Dwarkesh Patel, he argued the industry is moving from an age of scaling back to “an age of research,” where new ideas matter more than more compute.
  • Generalization is the core problem. He said current models do brilliantly on benchmarks yet generalize far worse than people, and that fixing this is central to SSI’s work.
  • Timelines. In the same interview he gave a range of roughly 5 to 20 years for systems that can learn as well as humans, longer than many CEO predictions. Compare others on our super intelligence timeline.
  • Showing, not telling. He has said that as AI becomes visibly more powerful, labs and governments will behave very differently, and that gradually showing what AI can do is important for safety.

SSI has not published its research methods, so outsiders can’t judge how close it is.

SSI vs other superintelligence labs

Lab Leader Products? Stated goal
Safe Superintelligence (SSI) Ilya Sutskever None Safe superintelligence, straight shot
OpenAI Sam Altman ChatGPT, API, devices AGI, then superintelligence
Meta Superintelligence Labs Alexandr Wang Meta AI across Meta apps “Personal superintelligence for everyone”
SpaceXSI (formerly xAI) Elon Musk Grok Superintelligence; see Musk and SI
Google DeepMind Demis Hassabis Gemini AGI, “responsibly”

What does “safe” mean at SSI?

Short answer: SSI hasn’t published a detailed definition. Its public statements describe safety as something to engineer into the system alongside its capabilities, not a set of rules added afterwards.

The founding statement uses an analogy: SSI wants safety in the sense of “nuclear safety,” not “trust and safety.” Content moderation, keeping a chatbot from saying offensive things, is not the goal. The goal is to make sure a system far more capable than people doesn’t cause catastrophic harm. That puts SSI squarely in the territory of the alignment problem: how to make an AI reliably pursue goals that are good for people even when it is smart enough to find loopholes.

The plan has three practical parts:

  • Advance capabilities and safety together, “making sure our safety always remains ahead.”
  • Avoid commercial pressure that could push the company to ship before it is ready.
  • Hire a small, very senior team rather than build a large organization.

What we don’t know is just as important. SSI has not published safety research, evaluation results, a responsible-scaling policy or a governance structure such as a nonprofit board. Labs that ship products, such as Anthropic, OpenAI and Google DeepMind, publish safety frameworks and system cards. SSI asks the public to trust its intentions without that evidence, at least for now.

Can a lab with no product win the race?

This is the central debate about SSI. Both sides have strong arguments.

Argument for SSI’s model Argument against
No product cycles: researchers spend all their time on the hard problem No revenue: every dollar of compute must come from investors
No pressure to release early to beat competitors Rivals spend tens of billions a year on compute; SSI has raised a few billion in total
A small team moves fast and keeps secrets well Small teams can be poached, as Gross’s exit showed
Sutskever’s track record (AlexNet, GPT series) attracts top talent No users means no real-world data on how its systems fail
If scaling has plateaued, new ideas, not money, decide the winner If scaling still works, the biggest spender wins

The last row is the crux. Sutskever’s bet only makes sense if he is right that the field has entered an “age of research.” If larger clusters keep producing better models, then OpenAI, Google, Meta and SpaceXSI, with their much larger compute budgets, have a structural advantage that a $32 billion startup cannot match.

Common confusions about “SSI”

The acronym causes plenty of mix-ups in search results:

  • SSI vs “SI.” “SI” is the term the US government adopted in 2026 for AI in general (details). SSI is a single company.
  • SSI vs Social Security. In the US, SSI usually means Supplemental Security Income, a benefits program. It is unrelated.
  • SSI stock. Safe Superintelligence is private. Public tickers or tokens using “SSI” belong to unrelated companies, and crypto tokens trading on the name are not affiliated with the lab.
  • SSI vs Meta Superintelligence Labs. Both use “superintelligence” in their names, but MSL is a division of Meta that ships products, while SSI is independent and ships none.

Our take: what to watch

SSI is the purest test of an idea many safety researchers hold: that the race to ship products is what makes AI dangerous. If a small, well-funded lab with no customers can reach the frontier first, that idea gains weight. If it can’t, the lesson may be that feedback from real users and revenue to pay for compute are not optional.

Three signals will tell you how SSI is doing:

  1. A first public result. A paper, a demo or a model, anything that lets outsiders measure progress.
  2. Another funding round. With no revenue, SSI must keep raising money to pay for compute. The price of the next round is a public vote on its progress.
  3. Talent flow. Gross’s departure showed that rivals can poach leaders. Watch whether key researchers stay.

Until then, Safe Superintelligence is the most expensive question mark in AI: a $32 billion lab whose work no one outside it has seen.

Enjoyed this? Get the SI Briefing

One short email a week: what changed on the road to super intelligence, and why it matters. No spam, unsubscribe anytime.

Frequently asked questions

What is Safe Superintelligence Inc.?

Safe Superintelligence Inc. (SSI) is an AI research company founded in June 2024 by former OpenAI chief scientist Ilya Sutskever, Daniel Gross and Daniel Levy. Its only stated goal is to build a superintelligence that is safe, without releasing products along the way.

Who owns Safe Superintelligence?

SSI is a private company owned by its founders, employees and investors. Reported investors include Andreessen Horowitz, Sequoia Capital, DST Global, SV Angel, Greenoaks Capital and Lightspeed, with Alphabet and Nvidia also reported as backers.

How much is Safe Superintelligence worth?

SSI was valued at about $32 billion in its 2025 funding, up from $5 billion in September 2024. No higher valuation has been confirmed by major outlets as of October 2026.

Does SSI have a product?

No. SSI has said it will not ship anything until it reaches safe superintelligence, so it has no public model, app or API and no reported revenue as of October 2026.

Is Safe Superintelligence a public company? Can I buy SSI stock?

No. SSI is privately held and not listed on any stock exchange. Tickers named SSI on public markets belong to unrelated companies.

5 Sources

  1. Safe Superintelligence Inc. — founding announcement (ssi.inc, June 2024)
  2. Safe Superintelligence Inc. (Wikipedia)
  3. Calcalist: SSI raises $2B at $32B valuation with no product yet (2025)
  4. OpenAI: Introducing Superalignment (July 2023)
  5. CIO: Ilya Sutskever launches Safe Superintelligence Inc. (June 2024)
Before you go

Super intelligence is moving fast.

Don’t miss what happens next

Get one short email a week with what changed and why it matters. No spam. Unsubscribe anytime.