Ilya Sutskever

Who they are

Ilya Sutskever is Co-Founder and CEO of Safe Superintelligence Inc. — co-founded OpenAI in 2015 and DNNResearch in 2012, the four-month Toronto lab acquired by Google that seeded modern deep learning.

Person

Ilya co-founded Safe Superintelligence Inc. in 2024, stepping into the CEO role in mid-2025 after co-founder Daniel Gross departed for Meta — launching it when the company was brand new and immediately capital-intensive, raising over $1B at a $5B valuation shortly after founding. His academic arc runs deep: Bachelor's then PhD in Computer Science at the University of Toronto under Geoffrey Hinton, a brief postdoc at Stanford, then a pivot to commercialisation — he co-founded DNNResearch in late 2012, a four-month Toronto lab that Google acquired almost immediately, bringing him in as a Research Scientist through 2015. He then co-founded OpenAI in 2015 and served as Chief Scientist for nearly a decade, leading the Superalignment team — an initiative aimed at aligning superintelligent AI — before departing in June 2024. He's co-founded four named ventures across his career: DNNResearch (acquired by Google), OpenAI (AI research lab), the Superalignment project at OpenAI, and Safe Superintelligence Inc. The through-line is a researcher who keeps betting bigger on the same question — what happens when AI surpasses human capabilities — and keeps building the institution he thinks is best placed to answer it. He appears on the Dwarkesh Patel podcast and the Lex Fridman podcast, the Eye On A.I. podcast, and gave a TED Talk in 2023 titled 'The exciting, perilous journey toward AGI'; his public voice consistently addresses scaling limits, alignment, and post-scaling research direction.

Company

The most recent development at SSI is Sutskever stepping into the CEO role in mid-2025, after co-founder Daniel Gross left to join Meta Superintelligence Labs — a departure that followed Meta's failed attempt to acquire SSI outright in the first half of 2025. Before that, SSI raised $2 billion in April 2025 in a round led by Greenoaks Capital Partners, with Andreessen Horowitz, Lightspeed, DST Global, Alphabet, and Nvidia participating, pushing the valuation to $32 billion — up from $5 billion at the September 2024 Series A. A subsequent $6 billion funding round pushed the valuation further; SSI has now raised approximately $3 billion across three rounds from 11 investors. SSI has a strategic partnership with Google Cloud for access to specialised TPUs. As of mid-2026, the company remains in a research-only posture — no product has been released, no timetable announced — with a team of roughly 20–27 people pursuing what it calls a 'straight shot' research program.

Market

SSI competes — philosophically more than commercially — against OpenAI, Anthropic, Google DeepMind, xAI, and Mistral AI, all of which commercialise intermediate AI products while SSI deliberately does not. Its distinguishing position is the refusal to ship anything until superintelligence is achieved safely, a bet that attracts regulatory tailwinds as governments increasingly favour companies with transparent safety protocols. The broader superintelligence race is marked by intense geopolitical competition and a lack of coherent global governance, with international calls for moratoria and concentration-of-power concerns creating both risk and legitimacy for a safety-first lab.

Network

Sutskever's closest professional circle at SSI is Daniel Levy, his remaining co-founder and a former OpenAI researcher. His formative research relationship is with Geoffrey Hinton, his doctoral mentor at Toronto and a long-running collaborator on deep learning. At OpenAI his key peers included Sam Altman (who recruited him to co-found it), Greg Brockman, Jakub Pachocki (who succeeded him as Chief Scientist), and Jan Leike (Superalignment co-lead).

Daniel Levy· Co-founder, Safe Superintelligence Inc. (former OpenAI researcher)
Daniel Gross· Former co-founder, SSI — departed to Meta Superintelligence Labs, mid-2025
Geoffrey Hinton· Deep learning mentor and research collaborator, University of Toronto
Sam Altman· CEO, OpenAI — recruited Sutskever to co-found OpenAI
Jakub Pachocki· Chief Scientist, OpenAI — succeeded Sutskever in the role
Jan Leike· Former Superalignment team collaborator, OpenAI
Greg Brockman· President, OpenAI — peer during Sutskever's tenure

How they likely show up

Long tenure at OpenAI (co-founder through ~9 years) followed by immediately founding SSI on the same core question → thinks in decade-long research arcs, not product cycles.
SSI deliberately has no product and no revenue timetable → comfortable operating in deep uncertainty; won't be rushed by commercial pressure.
Four founded companies across his career (DNNResearch, OpenAI, Superalignment, SSI) → high agency, consistently self-directs rather than joins someone else's mission.
Recurring podcast and TED appearances on scaling limits, alignment, and post-AGI governance → comfortable being a public intellectual, not just a researcher; likely responds well to substantive technical or philosophical engagement.
SSI team of 20–27 people at a $32B valuation → values extreme researcher density over organisational scale; probably allergic to process and overhead.
Publicly declared in the 2024 Dwarkesh interview that 'we're moving from the age of scaling to the age of research' → has a specific, stated thesis about where AI is going; will engage seriously with anyone who challenges or builds on it.

Conversation tips

→ Engage with his 2024 Dwarkesh thesis — 'the age of scaling is over, the age of research begins' — either to probe what that means for SSI's architecture decisions or to push back; he'll have views and will appreciate that you've heard it.
→ Ask about the Superalignment project specifically — he built and led it at OpenAI, it was dissolved after he left, and SSI is in many ways its successor. It's a thread with history.
→ Don't treat the lack of a product as a gap to fill — SSI's entire identity is the 'no product until superintelligence' bet. Treat it as a design choice worth understanding, not a missing feature.
→ Reference his Hinton lineage if you work in ML — the Toronto deep learning group is foundational to his identity and he'll appreciate knowing you understand where modern deep learning came from.
→ He's now CEO as well as Chief Scientist after Gross left — it's a recent and significant shift in role. Asking how that changes how he spends his time is a genuinely open question.

Toolbox

Openers

Open on the 2024 Dwarkesh interview line — 'we're moving from the age of scaling to the age of research' — it's his clearest public statement of SSI's research thesis and a sharp entry point into what he's actually building toward.
Reference the Meta acquisition bid that SSI rebuffed in early 2025 — a $32B offer turned down, followed by co-founder Gross leaving anyway. It's the most defining test of SSI's independence to date and signals a lot about how Sutskever is thinking about control and mission.
Bring up DNNResearch — the four-month Toronto lab co-founded in 2012 that Google acquired almost immediately. It's the earliest proof that he's been converting research into institutional bets his whole career, and most people skip it.

Discovery questions

You said in 2024 that we're leaving the scaling era and entering a research era — what does 'research' mean concretely for SSI's program, and how do you know when you've made progress?
After the Superalignment team at OpenAI was dissolved following your departure, what did that tell you about whether safety work can survive inside a commercially-driven lab?
With Daniel Gross gone and you now running both the research and the company, how do you think about the tension between keeping the team at 20-odd people and the scale of capital you've raised?

Avoid

Don't frame SSI's lack of a shipped product as a vulnerability or a question of 'when will you monetise' — the no-product posture is SSI's core thesis, not an oversight, and treating it as a gap will signal you haven't done the work.

Make it yours