Ilya Sutskever

Who they are

Ilya Sutskever is CEO and Co-Founder of Safe Superintelligence Inc. — co-founded DNNResearch in 2012, sold it to Google within four months, then co-founded OpenAI in 2015 before leaving to start SSI, the only lab whose sole mandate is building safe superintelligence.

Person

Ilya founded SSI in June 2024, days after departing OpenAI, when SSI had zero employees, zero revenue, and a single, deliberately narrow mandate. His academic arc ran from The Open University of Israel to a Bachelor's and then PhD in Computer Science at the University of Toronto — where Geoffrey Hinton was his mentor — followed by a brief postdoc at Stanford. He co-founded DNNResearch in Toronto in 2012; Google acquired it within four months, and he went to Google Brain as a Research Scientist through late 2015. He co-founded OpenAI in 2015 and served as its Chief Scientist for nearly a decade, leading the Superalignment team — an internal initiative to align superintelligent AI — before departing in May 2024. He's spoken extensively on the Dwarkesh Patel podcast (twice: 2023 and 2024), at TED in 2023, on Lex Fridman's podcast in 2020, and on Eye On A.I. — his consistent public argument is that we've moved from the age of scaling to the age of research, and that alignment is the defining problem of the field. The through-line across DNNResearch, OpenAI, and SSI is a pattern of founding organizations at the frontier of what's computationally possible, then staying long enough to see the implications.

Company

The most significant recent development at SSI is Nvidia's $5 billion investment in July 2026, which came after SSI showed Nvidia its secret research; SSI is migrating from Google TPUs to Nvidia GPUs and gaining access to Nvidia's Vera Rubin platform, a move that increases compute capacity by an order of magnitude. That follows a $2 billion round in April 2025 led by Greenoaks Capital, with participation from Andreessen Horowitz, Lightspeed, DST Global, Alphabet, and Nvidia, which valued SSI at $32 billion — up from $5 billion at its September 2024 Series A. In July 2025, co-founder Daniel Gross departed to join Meta Superintelligence Labs after Meta's acquisition bid was rebuffed; Sutskever stepped up from Chief Scientist to CEO at that point. SSI remains in a research-only posture as of mid-2026 — no product, no announced release timetable — operating from Palo Alto and Tel Aviv with a team of roughly 20–27 people.

Market

SSI competes in the frontier AI race against OpenAI, Anthropic, Google DeepMind, xAI, and Mistral AI, but occupies a structurally distinct position: every competitor ships commercial products; SSI does not, and has stated it won't until it achieves safe superintelligence. The broader industry faces mounting regulatory pressure — California laws like SB 7 and AB 1018 mandate AI transparency — and geopolitical dynamics around US-China chip competition shape compute access; SSI's safety-first positioning is seen as a tailwind in this regulatory environment, even as global AI governance frameworks remain fragmented.

Network

SSI's founding team was Daniel Gross (former Apple AI lead, now at Meta Superintelligence Labs) and Daniel Levy (former OpenAI researcher), both of whom Sutskever brought directly from his OpenAI orbit. His deeper research lineage runs through Geoffrey Hinton at Toronto, and his decade at OpenAI created close ties with Sam Altman, Greg Brockman, Jan Leike (Superalignment collaborator), and Jakub Pachocki, who succeeded him as OpenAI Chief Scientist.

Daniel Gross· Co-founder of SSI (departed to Meta Superintelligence Labs, mid-2025)
Daniel Levy· Co-founder of SSI, former OpenAI researcher
Geoffrey Hinton· Deep learning mentor and collaborator, University of Toronto
Sam Altman· CEO of OpenAI, recruited Sutskever to co-found OpenAI
Jakub Pachocki· Chief Scientist at OpenAI, succeeded Sutskever
Jan Leike· Former Superalignment team collaborator at OpenAI

How they likely show up

Serial founder pattern (DNNResearch → OpenAI → SSI) → thinks in institution-building arcs, not in quarterly delivery cycles; each move was a bet on a new frontier, not an incremental step.
SSI's explicit 'no product until superintelligence' posture → unusually high tolerance for deferring commercial validation; likely uncomfortable in meetings framed around near-term revenue or go-to-market.
Led the Superalignment team at OpenAI — a self-initiated internal safety research mandate — → capable of creating organizational structures around ideas he believes in, even within larger institutions.
Recurring public argument across Dwarkesh (2023, 2024), TED (2023), and Lex Fridman (2020) that we've moved from scaling to research → engages publicly when he has a substantive thesis, not for visibility; likely responds to intellectual provocation over flattery.
Rebuffed Meta's $32 billion acquisition bid in 2025 → strong mission conviction; not easily moved by headline valuations or institutional pressure.
Tiny, senior-only team (20–27 people) at a $32 billion valuation → values density of talent over headcount; likely impatient with process overhead and organizational bureaucracy.

Conversation tips

→ Engage him on the specific claim from the 2024 Dwarkesh interview — 'we're moving from the age of scaling to the age of research' — ask what that means for SSI's compute-heavy Nvidia partnership.
→ Don't frame anything around product timelines or commercialization; SSI's entire identity is that it won't ship until the problem is solved.
→ If you've read any of his public interviews rather than summaries of them, say so and cite a specific argument — he'll distinguish quickly between people who engaged with the ideas and those who scanned headlines.
→ The Superalignment team at OpenAI is worth asking about: he built it, it was dissolved after he left, and that arc likely carries weight for him.
→ Acknowledge the Nvidia shift explicitly if relevant — moving from Google TPUs to Nvidia's Vera Rubin platform is a significant infrastructure bet and a signal about where he thinks the research is heading.

Toolbox

Openers

Open on Nvidia's $5 billion investment in July 2026 and the TPU-to-GPU migration — SSI showed Nvidia its secret research to secure the deal, and the compute bet now underpins SSI's entire research program.
Reference the 2024 Dwarkesh Patel interview specifically: 'the age of scaling is over, we're in the age of research' — that's the thesis SSI was built on, and asking how that thesis is holding up in 2026 is a live question.
Bring up the Superalignment project at OpenAI — he founded it as an internal initiative on aligning superintelligent AI, it was dissolved after his departure, and SSI is in many ways its continuation outside OpenAI.

Discovery questions

The Nvidia investment came after SSI showed them internal research — what does it mean for SSI's model when a chip company is your largest single investor and compute partner?
You've argued publicly that scaling is no longer the primary driver — how has that claim aged given how much of SSI's current momentum is still compute-dependent?
With Daniel Gross now at Meta and the team at roughly 20–27 people, how do you think about the relationship between organizational size and research depth at this stage?

Avoid

Don't ask about the OpenAI board drama or frame his departure as a conflict story — he's moved on institutionally and SSI's identity is explicitly forward-looking; relitigating 2023–2024 internal politics won't open a productive conversation.

Make it yours