Sometimes a reorg says more about strategy than any press release. This is one of those cases: Google DeepMind has disbanded the team behind AlphaFold — the system that won the 2024 Nobel Prize in Chemistry. The researchers are moving into Gemini projects. And three of the most important minds are leaving entirely — for Anthropic.
What happened
As the Financial Times reported in late July, DeepMind reassigned most authors of the original AlphaFold paper to other roles over the past year. Nearly a quarter of the core authors have now left the company outright. Research VP Pushmeet Kohli confirmed a fundamental shift — away from the years-long strategy of focusing on single grand scientific challenges.
Nobel laureate John Jumper — whose move we covered back in June — didn’t go alone. Two more central AlphaFold researchers, Jonas Adler and Alexander Pritzel, followed him to Anthropic. One DeepMind employee described all three as “key, important, core members” whose collective departure sent real shockwaves through the company.
Why this is more than a personnel note
DeepMind is consolidating its resources around Gemini as a general-purpose platform — for search, cloud and scientific workloads at once. The era of standalone “grand challenge” teams looks to be over. It’s a defensible bet: one big model that does everything, instead of many specialized teams.
For Anthropic it’s a double win. First, talent almost nobody in the world has — people who proved that AI can deliver real scientific breakthroughs. Second, it lines up perfectly with Anthropic’s own direction: Claude Science, a workbench for biology and drug discovery, arrived only recently — the exact field AlphaFold once opened up.
My take
I find this remarkable — and a little wistful. AlphaFold was the prime example that AI doesn’t just write text but pushes hard science forward. That this team of all teams is being taken apart, while its minds move to a competitor, says a lot about the current balance of power. Anthropic is very deliberately stockpiling science talent right now. If that pays off, Claude could in a few years stand where AlphaFold began. Fascinating — and a signal I wouldn’t underestimate.
Sources: