Google DeepMind is navigating a difficult stretch. Last week, longtime CEO and company cofounder Demis Hassabis announced he was stepping back from the CEO role and into a new chair position. Day-to-day control passed to DeepMind's chief technology officer Koray Kavukcuoglu, who now reports directly to Alphabet CEO Sundar Pichai rather than holding a standalone DeepMind chief title. Minutes after that announcement, news broke that chief scientist Jeff Dean was leaving as well.
The back-to-back departures landed on top of an already difficult year for the lab. Gemini 3.5 Pro, unveiled with fanfare at Google I/O in May, has now blown through three release dates. Independent benchmarking firm Artificial Analysis told Fortune that Google's best shipped model, Gemini 3.6 Flash, currently trails Anthropic, OpenAI, xAI, Meta, and at least one Chinese lab on raw intelligence — a reversal from the brief but memorable stretch last year when Google actually topped the AI leaderboard.
The talent picture darkened further in June. In a single week, Google lost Gemini co-lead Noam Shazeer to OpenAI and AlphaFold co-inventor John Jumper to Anthropic. Two more AlphaFold veterans, Jonas Adler and Alexander Pritzel, followed Jumper out the door. Engineers cited aggressive poaching from cash-flush rivals, frustration at falling behind on coding benchmarks, and the pull of pre-IPO equity at competing firms.
Several engineers told Fortune's Beatrice Nolan that the reshuffle compounds a slow-motion shift of power from London, where DeepMind was founded, to Mountain View, where Google is headquartered. Their concern: that losing Hassabis means the longstanding firewall between DeepMind and its parent is breaking down. A Google spokesperson disputed that framing, saying London remains central and that DeepMind retains its research autonomy under the new structure.
Google is not without structural advantages. It still commands the chips, the cloud infrastructure, and the distribution that no AI-native rival can match — assets that could offset a weaker showing at the model layer. Just four years after Google's first AI 'code red,' however, DeepMind is again trying to prove it can move fast enough to keep pace, this time without Hassabis or Dean.
CEO Times take: The market has already voted on what happens when a research lab loses its founding scientific identity in the middle of a product race. Structural advantages — compute, cloud, distribution — are real, but they are table stakes, not a moat. What made DeepMind worth acquiring in the first place was the concentration of scientific talent and the autonomy to deploy it. If engineers are right that Mountain View is pulling the reins tighter, Google risks converting a world-class research engine into another product division. Capital rewards clear rules and clear accountability. The question now is whether Kavukcuoglu's direct line to Pichai accelerates decisions or simply centralizes them — and whether the engineers still inside DeepMind will wait long enough to find out.
