SSI is a private research lab led by CEO Ilya Sutskever and President Daniel Levy, working from Palo Alto and Tel Aviv. It sells no chatbot, API or model and publishes no research. Its sole planned product is the safe superintelligence it is trying to build, paid for by venture capital and a July 2026 NVIDIA compute partnership.
Safe Superintelligence Inc. (SSI) is a privately held American AI lab with one stated goal, a safe superintelligence, and no product planned before it. In July 2026 NVIDIA invested in SSI and opened its Vera Rubin systems to the lab, which the two companies say will raise SSI's compute tenfold.
Founded: 2024 · HQ: Palo Alto, CA, USA · Team: 20-30 · CEO: Ilya Sutskever · Funding: Disclosed: $1B (Sep 2024; NFDG, a16z, Sequoia, DST Global, SV Angel) and $2B (2025, led by Greenoaks Capital); NVIDIA invested in Jul 2026, amount undisclosed, reported by Bloomberg as $5B · Valuation: $32B (last disclosed, 2025 round led by Greenoaks)
About Safe Superintelligence Inc.
Ilya Sutskever announced Safe Superintelligence Inc, usually shortened to SSI, on X, naming Daniel Gross and Daniel Levy as his co-founders. Sutskever had left OpenAI a month earlier after years as its chief scientist, where he co-led the superalignment research. Gross had run AI efforts at Apple after Apple bought his startup Cue, and Levy had worked at OpenAI on optimization and large-scale training. SSI calls itself an American company with offices in Palo Alto, California and Tel Aviv, Israel, and its homepage still carries the founding line: SSI is our mission, our name, and our entire product roadmap.
Unlike other frontier labs, SSI has never shipped a product, an API, a chatbot, a model card or a research paper. Its one stated output is safe superintelligence itself, an AI system that exceeds human ability across essentially all cognitive tasks while staying reliably aligned with human intent. The company says it approaches safety and capabilities in tandem, pushing capability forward as fast as possible while keeping safety ahead, and it treats any intermediate commercial product as a distraction. There is no free tier, subscription or per-token rate to compare, because nothing is for sale.
Compute is where SSI has moved most since founding. Google Cloud agreed in 2025 to supply it with TPU capacity, and on July 27, 2026 SSI and NVIDIA announced a long-term partnership. Under that deal NVIDIA took an equity stake and gave SSI access to its Vera Rubin systems, which the two companies say will expand SSI's compute by an order of magnitude. NVIDIA's release says it entered the deal after getting rare access to SSI's closely guarded research, and it quotes Sutskever saying SSI has research worthy of scaling up.
The team is deliberately small and senior. Roughly 20 employees were reported around the March 2025 round, and SSI publishes no current headcount, revenue figure or financial statement. Its one leadership change came in mid-2025, when Meta tried to acquire the lab, was turned down, and hired Gross for its new superintelligence unit instead (see Meta AI).
Set against its peers, the trade-off is plain. Anthropic funds its safety work from Claude subscriptions and enterprise contracts, and OpenAI runs ChatGPT and the GPT-5 API; SSI carries none of that product overhead and none of that income. Google DeepMind has comparable TPU access plus direct routes into Google products. Well-funded labs such as xAI and Mistral AI release models while they research, whereas SSI has said it will release nothing before its namesake goal.
Who can ignore SSI for now: anyone choosing a model, an API or a vendor today. There is nothing to test, benchmark or buy, and with no papers or safety framework published, outside claims about its progress cannot be checked. The reason to keep watching is the combination of Sutskever's track record and a single deal meant to raise its compute tenfold; a first public research result would say a lot about how close the company believes it is. For labs whose products you can use today, browse the AI companies directory.
Mission
Building safe superintelligence (SSI) is the most important technical problem of our time.
Products
- Safe Superintelligence (unreleased) (Unreleased research program; SSI's sole stated product): https://ssi.inc
Links
Frequently Asked Questions
Can you buy Safe Superintelligence stock?
No. SSI is a privately held company with no stock ticker, and it has not announced any IPO, so its shares cannot be bought through a brokerage account (Motley Fool explainer, updated June 28, 2026). Its equity sits with the founders, staff and venture backers such as Andreessen Horowitz, Sequoia Capital, DST Global and Greenoaks. Two backers are listed companies, NVIDIA and Alphabet, but owning their shares gives only indirect exposure, since SSI is a very small part of either business.
Who founded Safe Superintelligence and who is the CEO?
SSI was founded on June 19, 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy, the day Sutskever posted the announcement on X. Gross left on June 29, 2025 and moved to Meta. In a July 3, 2025 message published on ssi.inc, Sutskever wrote that he is formally CEO and Levy is President, and NVIDIA's July 2026 partnership release still names the two of them as the company's leaders.
How much funding has Safe Superintelligence raised?
The company's updates page records a $1 billion raise on September 4, 2024 from NFDG, a16z, Sequoia, DST Global and SV Angel, at a reported $5 billion valuation. A $2 billion round led by Greenoaks Capital in 2025 set the last disclosed valuation of $32 billion. NVIDIA's July 2026 investment came with no stated amount; Bloomberg reported it as $5 billion, which would put total disclosed funding near $8 billion.
What are Safe Superintelligence's closest competitors?
The comparison is lopsided, because SSI sells nothing while its rivals run large product businesses. Anthropic and OpenAI pay for research from Claude and ChatGPT revenue, and Google DeepMind draws on Google's TPUs and products. For people rather than customers, the sharpest rival is Meta's superintelligence unit, which hired co-founder Daniel Gross in 2025.
What separates Safe Superintelligence from Anthropic?
Anthropic sells Claude to consumers and businesses and publishes model cards, research papers and a Responsible Scaling Policy. SSI has published none of those and takes no customers, so there is no product to compare and no public safety framework to audit. Someone picking an AI assistant today can only pick Anthropic; SSI matters mainly to people tracking where frontier talent and compute are going.
Top Alternatives
- Anthropic: Anthropic ships Claude and a public scaling policy; SSI ships nothing and publishes no framework yet.