Google DeepMind Disbands AlphaFold Team to Pivot Toward Large Models
Taylor Wilson
Google DeepMind has disbanded the core team behind its Nobel Prize-winning AI system AlphaFold, with most original authors reassigned or departed over the past year. This means → the lab is shifting wholesale from dedicated science-problem teams to a strategy built around its Gemini large language model.
What happened to the AlphaFold team?
Most original authors of the AlphaFold papers have been internally reassigned over the past year — to Gemini large-model projects, enzyme design, nuclear fusion, and genomics.
Some moved to Isomorphic Labs, Alphabet's drug-discovery subsidiary.
Nearly a quarter of the full-time Google DeepMind authors on the original paper have left the company entirely. In plain terms = the people who built AlphaFold have either changed direction or walked out the door.
Which key figures left?
Core AlphaFold researchers John Jumper and Jonas Adler first transferred to Google's internal "Code Strike" team focused on AI coding, then Jumper announced he was leaving to join Anthropic.
Adler and colleague Alexander Pritzel followed him to Anthropic.
An unnamed DeepMind employee called all three "pivotal members of the company." Their departure sent shockwaves internally. This reflects a broader talent shift — top AI researchers are moving from science-focused labs toward frontier large-model companies.
Why is DeepMind changing strategy?
DeepMind VP of Research Pushmeet Kohli told the Financial Times: the past nine years focused on "grand challenges — each project had a specific goal." That strategy has now evolved.
The new direction: build Gemini-powered systems that assist scientists and eventually automate parts of the scientific process. This means → DeepMind no longer assembles a dedicated team for each scientific problem; it is betting on one general-purpose model to span multiple fields.
At the same time, DeepMind must compete with OpenAI and Anthropic to develop frontier AI agents — a race for talent, compute, and market position all at once.
What does this reveal about the broader AI industry?
Frontier labs increasingly view large language models as tools to accelerate scientific research, not just chatbots. Anthropic this year launched Claude Science, a version designed specifically for biology and drug discovery.
OpenAI's head of research Mark Chen put it bluntly: "AI researchers want to work at a lab that is at the frontier, not chasing someone else." In plain terms = whoever has the strongest model attracts the best people.
DeepMind says it is "incredibly proud of the scientific legacy and global impact of AlphaFold," but the accelerating loss of its core team creates greater uncertainty about whether it can maintain its research edge in the large-model race.
Where did AlphaFold's legacy go?
AlphaFold development began at DeepMind in 2018. It used AI to predict the 3D structures of proteins — protein folding, the problem of knowing what shape a protein takes, is foundational to understanding life and designing drugs — and transformed biological research.
The breakthrough earned Jumper and DeepMind CEO Demis Hassabis the 2024 Nobel Prize in Chemistry.
The commercial pathway is now carried by Isomorphic Labs, spun out from DeepMind in 2021 and partnered with pharma giants Novartis and Eli Lilly. This means → AlphaFold's scientific output hasn't vanished; it has moved from a lab project into an independent company's commercial pipeline.
Content is for reference only, not financial advice.