Demis Hassabis Steps Down as Google DeepMind CEO, Shifts to AGI-Focused Chairman Role
Koray Kavukcuoglu takes over daily operations and Gemini development as Hassabis moves to chair DeepMind and chief scientist at Alphabet. The stock dropped 5% on the news.
Demis Hassabis is stepping down as CEO of Google DeepMind. He’s moving to Chairman of DeepMind and Chief Scientist of Alphabet, handing daily operations to Koray Kavukcuoglu, DeepMind’s former CTO and Google’s chief AI architect. Kavukcuoglu becomes Senior Vice President of Google DeepMind, reporting directly to Sundar Pichai.
Alphabet shares fell roughly 5% the day the reshuffle was announced, August 5.
Kavukcuoglu now owns Gemini model development, frontier AI research, and the Gemini app and developer teams — the parts of DeepMind that ship product and generate revenue. Hassabis keeps a narrower, higher-altitude mandate: AGI strategy and scientific applications of AI, with more time going to Isomorphic Labs, Alphabet’s AI drug-discovery unit that he continues to lead. Hassabis has said he believes “AGI is close at hand and that getting the next steps right is critical for humanity” — the stated rationale for the split.
The timing is not flattering. Gemini 3.5 Pro, DeepMind’s next flagship model, has slipped months past its original June target, and the reshuffle lands the same week Google is bleeding senior research talent. Chief scientist Jeff Dean and senior fellow Sanjay Ghemawat are both leaving after decades at the company to found an independent AI research outfit called Discovery Loop. That’s on top of ongoing departures to OpenAI and Anthropic that have quietly thinned DeepMind’s bench over the past year.
Splitting the CEO role into an operator (Kavukcuoglu) and a strategist (Hassabis) is a structure Google has used before — it’s roughly the Page/Schmidt playbook, and more recently how Microsoft runs its AI division under Mustafa Suleyman while Satya Nadella sets direction. The bet is that Kavukcuoglu, who has run Gemini’s technical roadmap for two years, can tighten shipping cadence while Hassabis spends his Nobel-laureate credibility on the research bets that take longer to pay off — better protein folding, better world models, the stuff that doesn’t fit a quarterly earnings call.
The risk is obvious: a CEO transition during a model-delay crisis and a talent exodus is exactly when you don’t want ambiguity about who’s accountable for shipping. Investors read the stock drop as skepticism, not confidence, that this fixes DeepMind’s execution problem before OpenAI’s GPT-5.6 line and Anthropic’s Opus 5 pull further ahead on developer mindshare.