Google DeepMind disbands Nobel-winning AlphaFold team

Google DeepMind has disbanded the team behind AlphaFold, its Nobel Prize-winning protein structure prediction system, marking a decisive pivot away from specialized scientific AI research toward its Gemini large language model platform, according to a report by the Financial Times.ft

A Gradual Unraveling

The dismantling unfolded over the past year through a series of personnel changes. Most of the original authors of the landmark AlphaFold paper have been reassigned to other projects, while nearly a quarter of those who remained as full-time employees have left the company entirely. Those still at DeepMind have been redirected to Gemini-related work or newer initiatives in enzyme design, nuclear fusion, and genomics. Others have moved to Isomorphic Labs, Alphabet's ↗ 0.16% drug discovery subsidiary.aibase

Pushmeet Kohli, DeepMind's vice president of research who founded and led its "AI for Science" team, acknowledged the strategic shift. His team had previously operated under a nine-year plan focused on tackling major scientific challenges with clear objectives. DeepMind, he said, no longer concentrates solely on single scientific problems like protein folding but is instead building Gemini-driven systems to assist scientists and eventually automate portions of the research process.aibase

Key Talent Departs for Rivals

The most visible consequence has been a talent exodus from the core team. Earlier this year, John Jumper — who shared the 2024 Nobel Prize in Chemistry with DeepMind CEO Demis Hassabis for AlphaFold's breakthrough — and another researcher named Adler joined an internal Google effort called "Code Strike," aimed at matching Anthropic and OpenAI in AI coding capabilities. Last month, Jumper announced his departure to join Anthropic, with Adler and AlphaFold colleague Pritzel following. A DeepMind employee described the three as "key, important, and core members" whose exits caused a stir internally.instagram+2

Broader Industry Implications

The restructuring reflects a wider pattern across major AI labs, where competitive pressure to build general-purpose AI agents and large language models is drawing resources away from domain-specific scientific tools. DeepMind is now directly competing with OpenAI and Anthropic for supremacy in AI agents — a contest that appears to have overtaken the lab's earlier mission of solving fundamental scientific problems one at a time.finance.biggo+1