# Safe Superintelligence **Safe Superintelligence Inc.** is [[Ilya Sutskever]]'s post-[[OpenAI]] research lab, founded June 19, 2024 on a single premise: build superintelligence safely, and ship nothing along the way. Two years in it has no product, no revenue, no published research, roughly fifty employees — and a valuation around $32B. It is the purest expression of the AI capital market's willingness to price a worldview rather than a business. That makes SSI the control case for the rest of the [[AI Sector]]. Every other frontier lab argues that commercialization is *necessary* to fund safety work — that you must ship to survive. SSI is the test of whether that claim is a constraint or an alibi. ## The Founding Premise Sutskever left OpenAI in May 2024, weeks after the Superalignment team he co-led with [[Jan Leike]] began to come apart, and six months after his vote to remove [[Sam Altman]] from the CEO seat. The SSI announcement on June 19, 2024 read as a direct rebuttal of what OpenAI had become: - **One goal, one product.** Safe superintelligence, developed as a single effort, with safety and capability treated as one engineering problem rather than a retrofit. - **Insulation by design.** No near-term commercial products, no revenue cycles, no management overhead — explicitly structured so that "our business model means safety, security, and progress are all insulated from short-term commercial pressures." - **Two offices.** Palo Alto and Tel Aviv. SSI is an Israeli-American company, which distinguishes its talent pipeline from the purely Bay Area labs. Co-founders: Sutskever, [[Daniel Gross]] (former head of AI at [[Apple]], [[Y Combinator]] partner, and half of the [[NFDG]] venture fund), and [[Daniel Levy]] (former OpenAI researcher). ## Funding Ladder | Date | Round | Amount | Valuation | Lead / notable backers | |------|-------|--------|-----------|------------------------| | Sept 2024 | Seed | $1B | ~$5B | [[Sequoia Capital]], [[Andreessen Horowitz]], [[DST Global]], SV Angel, [[NFDG]] | | Spring 2025 | Series A | ~$2B | $30–32B | [[Greenoaks]] leading; [[Alphabet]] and [[NVIDIA]] participating | | July 2026 | Strategic | $5B | not re-marked publicly | [[NVIDIA]] — see [[NVIDIA-Safe Superintelligence Investment]] | > [!warning] Contested numbers > The Series A is reported variously as March 2025 at $30B and April 2025 at $32B; cumulative-raise figures range from ~$3B (the sum of the two disclosed rounds) to ~$6B before NVIDIA's money arrived. The valuation is the number the coverage fixates on because it is the only number SSI has — there is no revenue, no user count, and no benchmark to anchor it. Treat the range as reporting variance, not as a correction. ## The Meta Approach and the Gross Departure In the first half of 2025, [[Meta]] tried to buy SSI outright. Sutskever declined. [[Mark Zuckerberg]] then bought the next best thing: in July 2025 Meta hired Gross and his NFDG partner [[Nat Friedman]] into [[Meta Superintelligence Labs]], absorbing a large slice of the NFDG portfolio in the process — an acqui-hire of the fund rather than the lab. Sutskever became CEO on July 3, 2025; Levy became president. His public note at the time was pointed: "We have the compute, we have the team, and we know what to do." This episode is the cleanest available demonstration of a pattern the KB tracks across the sector — when an incumbent cannot buy a competitor's mission, it buys the competitor's people. Compare the [[Scale AI]] transaction that brought [[Alexandr Wang]] to Meta, structured to obtain a founder without triggering a merger review. ## Compute: From Google TPUs to Vera Rubin - **April 2025:** [[Google]] Cloud announced a TPU arrangement supplying SSI's research compute — notable because [[Alphabet]] was simultaneously an investor and a competitor via [[Google DeepMind]]. - **July 27, 2026:** [[NVIDIA]] invested $5B and gave SSI access to its Vera Rubin platform, an increase in available compute of roughly one order of magnitude. Sutskever's stated rationale: "We have research that is worthy of scaling up, and having access to a big NVIDIA computer will let us do so." The shift from Google silicon to NVIDIA silicon moved SSI from one strategic patron to another. It also folded the sector's most conspicuously non-commercial lab into the circular-financing structure that critics — Michael Burry most loudly, calling NVIDIA "the Cisco of the bubble" — argue inflates the whole [[AI Sector]]: the chipmaker invests in the customer, and the customer spends the money on chips. ## The No-Product Thesis SSI has published no papers, released no models, and disclosed no benchmarks. The bet is a "straight shot" — go directly for superintelligence rather than climbing a ladder of shippable intermediate products. Sutskever's November 2025 framing supplies the intellectual case. He argued that the industry has left the **age of scaling** for an **age of research**: pre-training as practiced since GPT-3 has exhausted its easy returns, compute is no longer the binding constraint, and the next gains come from ideas rather than from another order of magnitude. If that is true, a fifty-person lab with no product obligations is structurally better positioned than a 4,500-person company shipping quarterly. If it is false, SSI is a $32B wager on a research direction nobody outside the building has been allowed to evaluate. That opacity is the substantive critique. A lab founded on the claim that safety requires transparency has produced less public safety research than any of its peers — see [[AI Safety]] and the [[AI Ethics and Safety Ecosystem]] for the community that has noticed. ## Political Footprint — Conspicuously Absent SSI is the frontier lab that does not appear in the KB's political-money layer. As of August 2026 it has no known federal lobbying registration, no PAC activity, and no visible presence in the [[Leading the Future]] network or the state-preemption fight that has consumed [[OpenAI]], [[Anthropic]], [[Meta]], and [[Andreessen Horowitz]] — see [[AI Political Spending Ecosystem]]. Sutskever has not taken a public administration-facing role of the sort [[David Sacks]] occupies. Two readings, and the KB should hold both: 1. **Consistency.** A lab that refuses commercial distraction refuses political distraction on the same logic. No products means no regulatory surface to defend. 2. **Delegation.** SSI does not need its own lobbyists while [[NVIDIA]] and [[Alphabet]] — both now investors — carry the sector's deregulatory water. Absence from the disclosure record is not the same as absence from the coalition. The second reading is the one worth tracking. SSI's capital table now runs through the two companies with the largest direct stake in an unregulated compute market. > [!note] Watch items > Any SSI federal lobbying registration or state filing · a first published paper or model · re-marking of the valuation after the NVIDIA round · additional Tel Aviv-side government or defense entanglement · whether Levy or Sutskever ever accepts a policy-facing appointment. ## See Also - [[Ilya Sutskever]] · [[Daniel Gross]] · [[Daniel Levy]] · [[Nat Friedman]] · [[NFDG]] - [[OpenAI]] · [[Anthropic]] · [[Meta Superintelligence Labs]] · [[Thinking Machines Lab]] · [[Google DeepMind]] · [[xAI]] - [[NVIDIA]] · [[Alphabet]] · [[NVIDIA-Safe Superintelligence Investment]] - [[AI Sector]] · [[Silicon Valley AI Ecosystem]] · [[AI Ethics and Safety Ecosystem]] · [[AI Political Spending Ecosystem]] - [[Superintelligence]] · [[AI Safety]] · [[Existential Risk]] · [[Scaling Hypothesis]] · [[AI Arms Race]] · [[Recursive Self-Improvement]] - [[Superintelligence (Book)]] — the Bostrom text the company is named against