Safe Superintelligence
Pre-product AI lab racing to build superintelligence directly, founded by Ilya Sutskever.
On the record
Checkpoint: Results expected from Nvidia-backed 10x compute ramp on Vera Rubin chips
The call
SSI holds billions in cash and zero shipped products, betting everything on Sutskever reaching superintelligence before rivals' shipped models make the wait pointless.
Safe Superintelligence is a company trying to build artificial general intelligence (AGI), the kind of AI that could do almost any task a human can do. Right now it has money and computing power, but no finished product or customers paying for anything. The problem is that OpenAI, Google, Meta, and Anthropic are all ahead of it: they have released products people use every day, customers giving them money and feedback to improve with, and they can pivot into AGI research whenever they want. The one thing that could change everything is if Safe Superintelligence releases real research or a working system that proves it is further along than everyone else thinks within the next two to three years, though the departure of key team members to competitors is a red flag that people inside may doubt the plan.
Why
01
AI pressure: OpenAI, Anthropic, Google DeepMind (shipping frontier models). Rivals ship revenue-generating models and gather real usage data daily while SSI has published nothing and sells nothing. Could AI do the core job? Yes, likely.
02
Bigger players: Meta Superintelligence Labs. Meta tried to buy SSI outright in 2025, got rebuffed, then hired away co-founder Daniel Gross anyway. Could a giant copy it? Yes, easily.
03
What protects it. Sutskever's reputation plus $3-6B in cash and fresh Nvidia and Google backing buy time without needing a product yet.
What would change this call
SSI publishes a paper, benchmark, or demo by mid-2027, tied to its Nvidia 10x compute ramp. Would show SSI has a verifiable capability lead, not just capital and reputation.
Another senior SSI researcher departs for Meta, OpenAI, Anthropic, or a spinout within 12 months. Would show the Sutskever-only talent base is eroding further.
Meta or another giant renews acquisition or licensing talks with SSI within 6 months. Would show rivals see SSI's independent path as unlikely to outrun them.
Another senior SSI researcher departs for Meta, OpenAI, Anthropic, or a spinout within 12 months. Would show the Sutskever-only talent base is eroding further.
Meta or another giant renews acquisition or licensing talks with SSI within 6 months. Would show rivals see SSI's independent path as unlikely to outrun them.
Key numbers
- $32B valuation in Nvidia/Alphabet round
- $5B Nvidia investment
- Zero revenue, zero shipped product or API
Named risks
- Sole load-bearing asset is Sutskever's personal reputation with no shipped product to fall back on.
- $32B valuation rests on zero revenue and no published research as of July 2026.
- Senior talent departures (Gross to Meta 2025-07, Papini to spinout 2026-02) concentrate risk on remaining leadership.
- Compute infrastructure shift from Google TPUs to Nvidia GPUs creates execution and integration risk.
- No externally verifiable capability signal exists to confirm internal research claims.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| OpenAI | losing to | OpenAI ships ChatGPT and full commercial suite; SSI has no product as of July 2026 |
| Meta Superintelligence Labs | losing to | Poached SSI co-founder Daniel Gross July 2025, attempted SSI acquisition H1 2025 |
| Anthropic | losing to | Commercializes Claude and enterprise APIs while SSI remains pre-revenue as of 2026-07 |
| Thinking Machines Lab | contesting | Shipped Inkling open-weights model July 2026, also secured Nvidia Vera Rubin deal March 2026 |
| Google DeepMind | losing to | Rated top-3 safety lab by FLI mid-2026 index alongside Anthropic and OpenAI; SSI unrated |
Questions people ask
- Is Safe Superintelligence profitable in 2026?
- No. As of July 27, 2026, Safe Superintelligence has generated zero revenue and has no commercial product, according to reporting on its Nvidia partnership announcement. The company has stated it does not intend to sell products in the near future, relying entirely on investor capital instead.
- Who owns Safe Superintelligence and what is it worth?
- Safe Superintelligence is privately held, led by CEO Ilya Sutskever. On July 27, 2026, Nvidia and Alphabet took equity stakes valuing the company at $32 billion, with Nvidia's stake reported at roughly $5 billion, following an earlier $30-32 billion round led by Greenoaks Capital in March 2025.
- Has Safe Superintelligence shipped any product yet?
- No. As of July 27, 2026, SSI has not published research papers, released an API, or shipped any commercial product since its June 19, 2024 founding. It describes itself as a straight-shot lab with one product still in development: safe superintelligence itself.
- Who is Safe Superintelligence's biggest competitor?
- Meta Superintelligence Labs is the most direct rival, having poached SSI co-founder Daniel Gross in July 2025 and previously attempted to acquire SSI outright in H1 2025. OpenAI, Anthropic, and Google DeepMind all ship commercial products SSI has not.
- Why did Nvidia invest billions in Safe Superintelligence?
- On July 27, 2026, Nvidia and Alphabet invested in SSI at a $32 billion valuation, with Nvidia's stake near $5 billion, to grant SSI access to Nvidia's Vera Rubin GPU platform and expand its compute roughly tenfold within 12 months, shifting SSI away from Google TPUs.
- Could AI replace Safe Superintelligence's business model?
- SSI's core product is frontier AI capability itself, so it is not threatened by AI automation but by faster-shipping rivals. OpenAI, Anthropic, Google DeepMind, and Meta Superintelligence Labs already have shipped products and more capital, competing directly for the same superintelligence milestone.
Timeline
- 2026-07-27Nvidia and Alphabet invest at $32B valuation; Nvidia stake reported near $5B, 10x compute plan set
- 2026-02-05Senior engineer Shahar Papini departs SSI to co-found AI-verification startup Attestable
- 2025-07Co-founder and CEO Daniel Gross departs to Meta Superintelligence Labs; Sutskever becomes CEO
- 2025-06Meta attempts to acquire SSI outright; rebuffed by Sutskever
- 2025-04Google Cloud partnership announced to supply TPUs for SSI research
- 2025-03SSI reaches $30-32B valuation in round led by Greenoaks Capital
- 2024-09SSI raises $1 billion from SV Angel, DST Global, Sequoia, a16z at $5B valuation
- 2024-06-19Safe Superintelligence Inc. founded by Sutskever, Daniel Gross, Daniel Levy
- 2024-05Ilya Sutskever leaves OpenAI following board dispute over Sam Altman ousting
Revision history
29 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Safe Superintelligence Inc. (2026-07-28)
- From Safe Superintelligence to Thinking Machines, everyone's saying no to Meta (2025-06-23)
- Nvidia and Alphabet Invest in Safe Superintelligence... (2026-07-27)
- Nvidia Invests $5 Billion in Ilya Sutskever's SSI, Shifting Startup From Google TPUs (2026-07-28)
- Nvidia invests in Sutskever's SSI to 10x its compute (2026-07-27)
- NVIDIA Backs Sutskever's AI Safety Lab With $5B and Vera Rubin Supercompute (2026-07-27)
- Ilya Sutskever's Safe Superintelligence raises $5 billion from Nvidia despite not yet... (2026-07-27)
- Ilya Sutskever's Safe Superintelligence engineer walks away to fix what AI breaks (2026-02-05)
- Ilya Sutskever's Safe Superintelligence Inc. and NVIDIA Announce Long-Term Strategic Partnership (2026-07-27)
- Ilya Sutskever's Safe Superintelligence Inc. and NVIDIA Announce Long-Term Strategic Partnership (2026-07-27)
- Thinking Machines Lab (2026-07-28)