John Jumper announced on X on June 19 that he is leaving Google DeepMind after nearly nine years and will join Anthropic following a break. Jumper led the team behind AlphaFold2, the protein-structure prediction system that has been used by more than two million scientists across 190 countries, and shared half of the 2024 Nobel Prize in Chemistry with DeepMind CEO Demis Hassabis for that work. Anthropic has not disclosed his title or team assignment. The move was confirmed by Bloomberg, CNBC, and TechCrunch on June 19–20.
The hire is the second senior departure from Google's AI org in 48 hours — Noam Shazeer, co-author of the 2017 transformer paper and co-lead of Gemini, left for OpenAI on June 18. What makes Jumper's move different is the destination's apparent strategy. Throughout 2026, Anthropic has been quietly building an AI-for-science arm: a virtual cell lab partnership with the Allen Institute and Howard Hughes Medical Institute announced in February, chemistry-focused model evaluations, and a public Briefing: AI for Science event scheduled for June 30. Jumper arrives at the head of an org that was already pointed at biology.
The competitive shape of frontier AI has changed in the last twelve months. Google, OpenAI, and Anthropic are no longer just shipping general-purpose models against one another — they are recruiting Nobel-class scientific leadership and building dedicated science divisions because the next obvious frontier for these systems is autonomous research. AlphaFold made structural biology routinely tractable; the bet at Anthropic appears to be that a science-grade model with the right scaffolding can do the same for drug discovery, materials, and protein design at industrial scale.
Takeaway for learners: if you are weighing what to study at the intersection of AI and the sciences, this hire is a signal. The labs are not just trying to win benchmark crowns — they are trying to build organizations capable of producing Nobel-Prize-quality science. That means demand for people who can translate between AI research and a domain — molecular biology, chemistry, materials, neuroscience — is going to be unusually high, and the work itself is going to look more like a research collaboration than a model-deployment pipeline.