Google’s DeepMind is having an identity crisis

Bea Nolan here, filling in for Allie. Google DeepMind has been having a difficult year. 

Last week, longtime Google DeepMind CEO and company cofounder Demis Hassabis announced he was stepping back from the CEO role and into a new chair role. 

Hassabis ceded day-to-day control to DeepMind’s chief technology officer Koray Kavukcuoglu—who now reports straight to Sundar Pichai, rather than holding a standalone DeepMind chief title. Just minutes later, news broke that chief scientist Jeff Dean was leaving too. 

The high-profile exits appeared to be just the tip of the iceberg in DeepMind’s tough year. 

Gemini 3.5 Pro, unveiled with fanfare at I/O in May, has now blown through three release dates. Independent benchmarking company Artificial Analysis also 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.

Several engineers told me the reshuffle compounds a slow-motion pull of power from London, where DeepMind was founded, to Mountain View, where Google is headquartered. They worry losing Hassabis means the longstanding firewall between DeepMind and parent Google is breaking down. (A Google spokesperson disputes that framing, saying London remains central and that DeepMind keeps its research autonomy under the new structure.)

The company is also struggling to keep hold of some of its key talent. In a single week in June, 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 blame the exodus on aggressive poaching from cash-flush rivals, frustration at falling behind on coding benchmarks, and engineers chasing pre-IPO equity before rivals go public. 

None of this means Google is out of the race. It still has the chips, the cloud, and the distribution that none of its AI-native rivals can touch—structural advantages that could offset a weaker showing at the model layer. But, just four years after Google’s first AI code red, DeepMind is again trying to prove it can move fast enough to keep pace. But this time, it will be doing so without the two figures (Hassabis and Dean) most closely associated with its scientific identity.

Read my whole story here.

See you tomorrow,

Beatrice Nolan
X: @beafreyanolan
Email: [email protected]

Submit a deal for the Term Sheet newsletter here.

Joey Abrams curated the deals section of today’s newsletter. Subscribe here.

This story was originally featured on Fortune.com