Google DeepMind has spent the past year quietly dismantling the team behind AlphaFold, the protein-structure model that earned the lab a Nobel Prize in 2024. Most of the scientists who co-authored the original AlphaFold paper have been moved onto other projects, and nearly one in four of them have left the firm altogether, according to the Financial Times. DeepMind confirmed the changes and said the work is being absorbed into a broader push built around Gemini, its flagship model line. The Next Web compiled the account from that FT reporting alongside DeepMind’s own statements.
DeepMind research vice president Pushmeet Kohli described the shift to the Financial Times: “Our strategy over the last nine years has been to focus on grand challenges. The strategy has evolved.” In practice that means fewer standalone teams chasing a single scientific problem and more Gemini-powered systems built to assist researchers or automate stretches of the work itself. AlphaFold, which predicts the three-dimensional shape of proteins in minutes and now underpins drug discovery worldwide, was the proof that the old approach could work.
Former AlphaFold staff scattered widely. Some moved onto other Gemini teams, others into enzyme design, nuclear fusion research and genomics work, The Decoder reported. A handful joined Isomorphic Labs, the Alphabet-owned drug discovery spinout that AlphaFold made possible in the first place.
The most consequential departures left DeepMind altogether. DeepMind chief Demis Hassabis and John Jumper jointly received the 2024 Nobel Prize, but Jumper left the lab for Anthropic in June. Jonas Adler and Alexander Pritzel, two more scientists central to the original AlphaFold work, left for the same company soon after. One DeepMind employee described the trio to the Financial Times as “instrumental, important, core members” whose departure “sparked surprise internally.”
The three now work at a company that just launched Claude Science, an Anthropic workbench built for the same biology and drug-discovery problems AlphaFold was designed to solve. The lab describes itself as “incredibly proud” of what the AlphaFold project achieved.
AlphaFold is close to the strongest argument available that patient, curiosity-driven science pays off inside a corporate lab. It ran for years without a product deadline, then produced a Nobel Prize and a drug-discovery industry now built on its output. Redirecting that team toward Gemini is a statement about what DeepMind’s parent company rewards today, and it is not multi-year science with no near-term revenue line.
The lesson extends past DeepMind. Any company running both a long-horizon research lab and a shipping chatbot business will eventually face this allocation question, and the AlphaFold breakup shows which side wins once a model race is underway. Anthropic’s hiring of Jumper, Adler and Pritzel suggests a rival can buy the payoff of that patience without funding the decade of research that produced it. For anyone weighing whether to bankroll open-ended research inside a product company, DeepMind’s answer is that grand challenges get reorganized the moment a faster-moving bet needs the headcount.
This account is based on The Next Web’s July 29, 2026 report, which cited original reporting from the Financial Times along with coverage from The Decoder and Engadget.
