Nobel laureate John Jumper leaves Google DeepMind for Anthropic — a signal Anthropic is going after AI for science

What’s confirmed

Per Jumper’s own X post, Hassabis’s reply, and TechCrunch:

• Who: John Jumper, AlphaFold lead at Google DeepMind, 2024 Nobel laureate in Chemistry (shared with Demis Hassabis) for AlphaFold — the system that predicts 3D protein structures from genetic sequences.
• What: leaving DeepMind for Anthropic after nearly nine years; taking time to “recharge” first.
• His words: Hassabis “took a real chance letting me lead the AlphaFold team just six months after finishing my PhD”; called DeepMind “a special place.”
• Hassabis’s response: thanked Jumper for “an extraordinary partnership,” noting AlphaFold “changed the world.”
• The pattern: same week as Noam Shazeer’s departure to OpenAI — two of DeepMind’s most prominent figures leaving for the two biggest rivals at once.
• Not confirmed: Jumper’s specific role or mandate at Anthropic.

Why this matters

Three reads.

1. This is the clearest signal yet that Anthropic wants to lead AI-for-science. Anthropic is best known for Claude’s writing and coding strength and its safety posture. But the threads have been there: Mythos was explicitly aimed at biomedical researchers (with biology safety classifiers), and the Fable 5 launch touted ~10× protein-design acceleration. Hiring the name in AI-for-biology — the person who led AlphaFold — turns that subtext into strategy. Jumper doesn’t join a lab to do incremental work; his presence says Anthropic intends to compete with DeepMind on scientific AI, the domain DeepMind has owned. For a company heading toward a public listing, “we’re not just a chatbot company, we’re a science company” is a powerful expansion of the story.

2. The same-week double departure is a real signal about Google. One marquee exit is noise; two founding-era figures — the Transformer co-inventor and the AlphaFold Nobel laureate — leaving for OpenAI and Anthropic in the same week is a pattern worth naming. It lands while Gemini 3.5 Pro has slipped past its promised June launch and Google is pushing developers off tools like Gemini CLI. None of that means Gemini is in trouble — it remains the clear #2 chatbot and DeepMind is enormous under Hassabis — but talent flows are a leading indicator, and right now the arrow points away from Google toward its two rivals.

3. The Hassabis irony underlines the stakes. Jumper shared his Nobel Prize with Hassabis, his boss — and is now leaving Hassabis’s DeepMind for Anthropic. When a co-laureate departs for a competitor, it’s not about comp alone; it’s about where someone believes the most important work will happen next. That Jumper bet on Anthropic is, in itself, a data point about Anthropic’s scientific trajectory that no press release could buy.

What it means for Claude and the AI-tools landscape

For Claude users today: nothing changes. This is a hire, not a product. Claude’s current strengths — writing, coding, reasoning — are unaffected, and the Claude review stands.

For the medium term: expect Anthropic to invest visibly in scientific and biological AI. If you work in research, drug discovery, computational biology, or adjacent fields, Anthropic is signaling it wants Claude (and successor models) to be serious tools for you — a space where Gemini has had a structural edge thanks to DeepMind’s science work. Watch for science-focused Claude features and research partnerships over the next year.

For the competitive picture: the three-way race (ChatGPT, Claude, Gemini) is no longer just about general chat and coding. Anthropic adding Jumper, OpenAI adding Shazeer, and both building custom silicon shows all three labs broadening — talent, science, and infrastructure — as they head toward public markets. For buyers, that breadth is good: more capable models across more domains, from more credibly-resourced labs.

The honest caveats

His role is unconfirmed. Anthropic hasn’t announced what Jumper will do, and he said he’s taking time off first. “Anthropic is going after AI-for-science” is a well-supported inference from his expertise and Anthropic’s existing biomedical threads — but it is inference, not an announced mandate.

One hire doesn’t build a division. A Nobel laureate is a powerful signal, but scientific AI at AlphaFold scale takes teams, data, and years. Treat this as a statement of intent, not a shipped capability.

Talent moves are noisy. People change labs for many reasons — scope, autonomy, conviction, timing, comp. The “Google is bleeding talent” narrative is real as a pattern this week, but two departures don’t determine the trajectory of an organization DeepMind’s size.

Bloomberg context, lightly held. Reporting noted Jumper had been involved in Google’s coding-tools work (which Google has struggled to commercialize); how much that motivated the move is unknown and we won’t speculate beyond noting it.

What it changes for Pick Right readers

Nothing about which AI tool to use today. But it sharpens a forward-looking thesis worth tracking: Anthropic is expanding from “the quality chatbot and coding company” toward AI-for-science, and it just hired the field’s most decorated figure to get there. If your work touches biology or research, that’s a reason to watch Anthropic’s roadmap closely. And the broader signal — two of DeepMind’s biggest names leaving for OpenAI and Anthropic in one week — is the kind of talent-flow data point that often precedes shifts in who leads.

For the connected threads, see the Noam Shazeer → OpenAI move, the Claude review, the Gemini review, the Fable 5 + Mythos 5 launch (the biomedical angle), the Anthropic S-1 filing, the OpenAI Jalapeño chip, the full DeepMind talent exodus + Gemini 3.5 Pro delay, the Claude Science workbench launch (this thesis, shipped as a product), and the best AI chatbots guide.

Questions or corrections? Email Pick Right. Want the full list? See all news.