Edited by 
Short answer
Safe Superintelligence Inc. (SSI) is an AI research company founded in June 2024 by Ilya Sutskever, OpenAI's former chief scientist, with Daniel Gross and Daniel Levy. Its only stated goal is to build superintelligence, AI far smarter than humans, that is safe, and it says it will release no products until then. SSI has offices in Palo Alto and Tel Aviv, was valued at about $32 billion in 2025, uses Google Cloud TPUs, and in July 2026 announced a compute partnership with NVIDIA that reportedly includes a $5 billion investment. The term "safe superintelligence" also describes the wider goal of making superhuman AI that stays aligned with human interests.
What is Safe Superintelligence Inc.?
Ilya Sutskever announced SSI on 19 June 2024, a month after leaving OpenAI, where he had co-founded the company and led its research. The announcement called SSI "the world's first straight-shot SSI lab, with one goal and one product: a safe superintelligence."
That framing is the company's defining feature. Most AI labs, including OpenAI, Anthropic and Google DeepMind, fund their research by selling products such as chatbots and APIs. SSI says it is deliberately "insulated from short-term commercial pressures" and free of "management overhead or product cycles", so that safety and capability can be developed together without the pressure to ship.
Who founded SSI?
- Ilya Sutskever: co-creator of AlexNet, the 2012 image-recognition network that kicked off the deep learning era, co-founder and former chief scientist of OpenAI, and co-lead of its Superalignment team. He became SSI's CEO in July 2025.
- Daniel Gross: former head of AI at Apple and an investor. He left SSI in July 2025 to join Meta Superintelligence Labs.
- Daniel Levy: a former OpenAI researcher who worked on optimization.
The company kept its team small at first, with about 20 employees in early 2025 and around 50 by mid-2025, and hires mainly researchers and engineers.
SSI funding and valuation timeline
| Date | Event |
|---|---|
| June 2024 | SSI founded by Sutskever, Gross and Levy |
| September 2024 | Raised $1 billion at a $5 billion valuation from investors including Andreessen Horowitz, Sequoia Capital, DST Global and SV Angel |
| Early 2025 | New round led by Greenoaks; later reports put the valuation at about $32 billion after a further $2 billion raise, with Alphabet and NVIDIA among backers |
| April 2025 | Google Cloud agreed to provide Tensor Processing Units (TPUs) for SSI's research |
| July 2025 | Daniel Gross left for Meta; Sutskever became CEO after turning down a Meta acquisition approach |
| July 2026 | Long-term partnership with NVIDIA giving SSI access to the Vera Rubin platform, expected to raise its compute "by an order of magnitude"; Bloomberg reported a $5 billion NVIDIA investment |
A $32 billion valuation for a company with no product and no revenue is unusual. It reflects investors' bet on Sutskever's track record and on the size of the prize if superintelligence is reached.
What hardware does SSI use?
Like every frontier lab, SSI's progress depends on compute. It first trained on Google Cloud TPUs, Google's custom AI chips; read TPU vs GPU for how they differ from NVIDIA GPUs. The 2026 NVIDIA deal adds access to Vera Rubin, NVIDIA's generation after Blackwell, built around Rubin GPUs, Vera CPUs and NVLink rack-scale systems. Read NVIDIA Rubin explained, and what is an AI data center for the kind of facility this compute lives in.
What does "safe superintelligence" mean?
Superintelligence means an AI that greatly exceeds human ability across virtually every field, not just in narrow tasks like chess. "Safe" adds the requirement that such a system reliably does what humanity actually wants and does not cause catastrophic harm, even though it would be smarter than the people checking it.
Researchers usually break the safety problem into a few parts:
- Alignment: making sure the AI's goals match human intentions, not a proxy that looks right in training.
- Scalable oversight: finding ways for humans, helped by weaker AI, to supervise a system smarter than they are.
- Robustness: keeping behaviour safe in situations the system was never trained on.
- Control: being able to monitor, correct and stop a system if something goes wrong.
Sutskever has compared the goal to nuclear safety rather than "trust and safety" content moderation: safety built into the system from the start, not filters added afterwards. SSI has not published details of its technical approach.
SSI vs other superintelligence efforts
| Lab | Approach | Products |
|---|---|---|
| Safe Superintelligence (SSI) | One goal, safe superintelligence; research only | None |
| OpenAI | Aims for AGI that benefits humanity; funds research through ChatGPT and its API | ChatGPT, API, Sora |
| Anthropic | Safety-focused research company funding work through Claude | Claude, API |
| Google DeepMind | Research arm of Google; models ship in Google products | Gemini |
| Meta Superintelligence Labs | Meta's 2025 effort to build "personal superintelligence" | Meta AI, Llama models |
Will SSI succeed?
Nobody knows. SSI's advantages are focus, a famous research leader and now access to some of the largest compute available. Its risks are the same as its strengths: with no product, it has no revenue and no public record of results, and competitors that ship products also gather huge amounts of user feedback and money for compute. Its first public output, whenever it comes, will be watched closely.
For the bigger picture of where AI is heading, read what is superintelligence, AGI vs ASI and how AI works.
Related: should superintelligence be banned? and AI safety in Australia.
Frequently asked questions
Who owns Safe Superintelligence?
SSI is a private company. Its founders and employees hold shares alongside investors including Andreessen Horowitz, Sequoia Capital, DST Global, Greenoaks, Alphabet and NVIDIA.
Does SSI have a product?
No. SSI says it will not release products until it has built safe superintelligence. It has no chatbot, API or public model.
Is SSI the same as "superintelligence"?
No. SSI is a company named after the goal. Superintelligence, often shortened to ASI, is the general idea of AI that surpasses humans across almost all domains. Read what is superintelligence.
Why did Ilya Sutskever leave OpenAI?
Sutskever was one of the OpenAI board members who voted to remove Sam Altman as CEO in November 2023, a decision reversed within days. He left OpenAI in May 2024 and founded SSI a month later.
Can I invest in SSI?
Not directly. SSI is privately held and not listed on any stock exchange.
Sources and further reading
- Safe Superintelligence Inc., official website
- TechCrunch, SSI partners with NVIDIA to scale its AI research (July 2026)
- Wikipedia, Safe Superintelligence Inc.
- Maginative, SSI raises another $2B at a $32B valuation
Leave a Reply