Google is restructuring the leadership of its flagship AI lab in a bid to accelerate Gemini model releases and close a widening performance gap with OpenAI and Anthropic. Demis Hassabis, who co-founded DeepMind and has led it since Google's 2014 acquisition, will become chair of Google DeepMind and chief scientist of Alphabet. Koray Kavukcuoglu, the lab's longtime chief technology officer, steps up as senior vice president of Google DeepMind, reporting directly to Sundar Pichai. The changes, announced in internal memos from both Pichai and Hassabis, also mark the departure of Jeff Dean after a 27-year career at Google. Dean will co-found a public benefit corporation called Discovery Loop with fellow senior fellow Sanjay Ghemawat.
The shake-up arrives at a pivotal moment. Google's Gemini 3.5 Pro has been delayed for months while competitors shipped frontier models that captured developer mindshare. Anthropic's Mythos family and OpenAI's GPT-5.6 both landed major capability jumps in 2026. Meanwhile, Google's own Gemini 3.1 Pro launched in February but failed to move the frontier forward. An internal "Code Strike" team was assembled to bolster coding performance, a domain where rivals are "miles ahead" according to Morningstar analyst Malik Ahmed Khan. That team has been working on synthetic data pipelines, reinforcement learning from execution feedback, and architecture tweaks aimed at narrowing the gap on code generation benchmarks.
Leadership shift signals execution priority
Kavukcuoglu's promotion reads as a clear signal that Google wants faster, more predictable model releases. The DeepMind veteran joined in 2012, before the Google acquisition, and helped pioneer breakthroughs including WaveNet and DQN. Over the past year he had already been taking on broader responsibilities from Hassabis, directing model development and presenting major Gemini launches.
"Expect faster releases, more experimentation and an emphasis on developing to a product roadmap," said Ray Wang, principal analyst at Constellation Research. Nick Patience, AI lead at the Futurum Group, added that Kavukcuoglu's first step must be shipping Gemini 3.5 Pro and then proving it wasn't a one-off by maintaining a predictable cadence. "Beyond shipping a model that competes at the frontier, he needs to rebuild the coding and pretraining expertise that walked out the door," Patience told CNBC.
The departures have been notable. Noam Shazeer, a Gemini co-lead, left for OpenAI. John Jumper, Nobel laureate for AlphaFold, moved to Anthropic. Hassabis himself acknowledged the tension between long-horizon research and daily operations. "We have to accelerate all this work and stay focused on the AI frontier," Pichai wrote in his memo. "At the same time, there's never been a more important moment to shape what comes next for AGI and science."
Hassabis pivots to AGI strategy and Isomorphic Labs
Hassabis isn't leaving the fold. As chair of Google DeepMind and chief scientist of Alphabet, he will advise across models and research while devoting more time to what he calls his "life's work" -- ensuring AGI develops safely. He will also continue leading Isomorphic Labs, the drug-discovery spinout that applies AlphaFold's protein-folding breakthroughs to pharmaceutical development.
In his own note to staff, Hassabis wrote that he feels AGI is "close at hand" and that getting the next steps right is critical for humanity. The language mirrors his recent public writing about standing in "the foothills of the singularity." Analysts say the split role lets him focus on high-level strategy while Kavukcuoglu handles the operational grind of model development, product integration, and developer relations.
Ben Wood, chief analyst at CCS Insight, told CNBC the goal will "undoubtedly be to close the gaps with Anthropic and OpenAI in some areas." He added that Kavukcuoglu's promotion "feels like a move designed to adjust the focus of Google's efforts away from academic projects and more toward a stronger focus on improving frontier performance and improving the toolset for developers."
Jeff Dean exits to launch Discovery Loop
The third major move is Jeff Dean's departure after 27 years. Dean and Ghemawat built the infrastructure that powered Google's early search dominance and later the neural-network revolution. Their new venture, Discovery Loop, is described as an independent public benefit corporation focused on "accelerating discoveries in ML, science and engineering." Alphabet is a founding investor and cloud partner.
Dean's exit underscores a broader pattern: the generation of researchers who built modern AI's foundations are increasingly spinning out to pursue ambitious, less constrained agendas. Google retains a financial stake and cloud relationship, but the day-to-day collaboration will shift to a research-framework model.
Market reaction and what comes next
Alphabet shares fell roughly four percent on the announcement, reflecting investor anxiety about whether the new structure can deliver the consistent frontier performance that has eluded Google in 2026. Cloud revenue grew 82 percent year-over-year to $24.8 billion in the latest quarter -- outpacing AWS and Azure -- but the AI narrative still centers on model leadership.
Gemini app usage has hit 950 million monthly users, and Gemma open models have surpassed 900 million downloads. Those distribution metrics are strong. The question is whether Kavukcuoglu can translate distribution into the kind of developer loyalty that OpenAI's API ecosystem and Anthropic's Claude platform now command.
For now, the lab enters a new chapter with a leadership team that blends DeepMind's research DNA with a mandate for faster execution. Whether that combination produces the next frontier breakthrough will define Google's AI trajectory for the next several years.
The timing of the reshuffle also reflects mounting pressure from outside investors. Venture capital firms have been betting heavily on Anthropic and OpenAI, with Anthropic recently closing a $14 billion round at a higher valuation than Google's own AI investments. Meanwhile, the European Union's AI Act is pushing Google to be more transparent about its safety testing processes, adding regulatory overhead to an already complex release schedule. These external forces compound internal challenges around talent retention and product velocity.
According to a Google spokesperson, the company remains committed to pushing on agents, coding, robotics, and world models as the path to AGI. The spokesperson also disputed the characterization that Google was distracted from its mission, pointing to recent releases across robotics, video, and computer use.
The coming months will test whether a tighter product focus can coexist with the kind of blue-sky research that produced AlphaFold, AlphaGo, and the transformer architecture itself. Hassabis will watch from the chair. Kavukcuoglu owns the roadmap.
Read Sundar Pichai's memo on the Google blog
For more on AI infrastructure trends, see our coverage of AI and Semiconductors.