DeepMind dismantles AlphaFold team; Jumper, Adler, Pritzel join Anthropic

Google DeepMind has dismantled the research team behind AlphaFold, the protein-structure prediction system that earned the 2024 Nobel Prize in Chemistry. According to the Financial Times, most of the original AlphaFold researchers have been reassigned to other projects over the past year, and nearly a quarter of the full-time DeepMind authors of the original AlphaFold papers have left the company altogether.

The reassignments scatter the team across multiple initiatives. Some researchers are now working on Gemini (Google's large language model)-powered systems, while others have moved into enzyme design, nuclear fusion, and genomics. A subset transferred to Isomorphic Labs, an Alphabet subsidiary focused on drug discovery that was spun out of DeepMind in 2021 and maintains partnerships with pharmaceutical companies Novartis and Eli Lilly. The most visible departures, however, are to rival company Anthropic: John Jumper (vice president and engineering fellow at DeepMind), Jonas Adler, and Alexander Pritzel have all announced moves to Anthropic. Jumper and Adler had been previously reassigned to Google's internal "Code Strike" team, which aims to improve the company's AI coding capabilities as it competes with Anthropic and OpenAI. One unnamed DeepMind employee told the FT that all three had been "instrumental, important, core members" and that their departures had sparked internal surprise.

The restructuring represents a sharp departure from the strategy that built DeepMind's reputation. AlphaFold, developed starting in 2018, defined a nine-year approach of assembling dedicated teams around grand scientific challenges. Pushmet Kohli, vice president of research at Google DeepMind, told the FT: "Our strategy over the last nine years has been to focus on grand challenges . . . a concrete goal every project is focused on. The strategy has evolved." Rather than organizing researchers around single scientific problems, DeepMind is now building Gemini-powered systems designed to assist scientists and eventually automate parts of the scientific process. Simultaneously, the lab is competing with OpenAI and Anthropic to develop frontier AI agents.

DeepMind stated it remains "incredibly proud" of its scientific heritage and AlphaFold's global impact, adding that the work "inspired the launch of Isomorphic Labs with the mission to reimagine drug discovery." Anthropic has also entered the space, having launched Claude Science for biology and drug discovery earlier this year. The departures underscore how competition for top AI talent—particularly researchers with breakthrough credentials—is intensifying among leading labs, and how strategic priorities at DeepMind are shifting from deep scientific inquiry to broader AI systems that compete in the frontier model space.

DeepMind's restructuring reflects a broader strategic reorientation within Google's AI labs. For nine years, the organization built its reputation on what VP Pushmeet Kohli calls the "grand challenges" approach—assembling dedicated teams around concrete, focused scientific goals like protein folding. That model produced AlphaFold, which earned John Jumper and CEO Demis Hassabis the 2024 Nobel Prize in Chemistry and became the defining example of DeepMind's scientific ambition. The dismantling of that team, however, signals that the lab's leadership now sees greater value in building Gemini-powered systems that can work across scientific domains and in developing frontier AI agents to compete with OpenAI and Anthropic.

The departure of three "instrumental, core members" to Anthropic—a direct competitor in the frontier AI space—adds competitive sting to the restructuring. Jumper and Adler had been moved internally to Google's "Code Strike" team, suggesting the company recognized their value for AI coding, but that pivot was short-lived. The loss is particularly significant because it removes from DeepMind some of the researchers most visibly associated with the lab's greatest recent achievement, potentially signaling to the research community that breakthrough-focused science teams at DeepMind have less strategic weight than they once did.

No comments yet. Be the first to share your thoughts!