DeepMind Disbands Its Nobel-Winning AlphaFold Team
A year after AlphaFold won the Nobel Prize, its team no longer exists: researchers reassigned to Gemini, John Jumper gone to Anthropic. What the breakup says about who funds deep science in the AI era.
Less than a year after AlphaFold won Google DeepMind a Nobel Prize in Chemistry, the dedicated team behind it no longer exists. According to reporting first published by the Financial Times and confirmed across Engadget and others, most authors of the original AlphaFold paper have been reassigned over the past year, nearly a quarter of the core full-time authors have left the company, and the protein-folding work has been folded into DeepMind's broader Gemini push.
The personnel losses aren't minor. John Jumper — the Nobel laureate who led AlphaFold — left for Anthropic in June, followed by two more core AlphaFold researchers. DeepMind research VP Pushmeet Kohli's explanation to the FT was as close to a strategy memo as a single sentence gets: the nine-year approach of attacking grand scientific challenges "has evolved."
What actually changed
It's worth being precise, because "Google shuts down AlphaFold" headlines overstate it. AlphaFold the product isn't going anywhere — the protein structure database remains available and underpins drug-discovery work worldwide. What's gone is AlphaFold the organizational model: a dedicated, long-horizon team pointed at one hard scientific problem until it cracks.
The replacement model is Gemini-centric. DeepMind says protein and structural biology work continues inside wider scientific programs — enzyme design, genomics, fusion — increasingly built around the idea of an automated "AI scientist" powered by its frontier models, rather than bespoke systems like AlphaFold.
Why a Nobel Prize didn't protect the team
The uncomfortable logic is straightforward. AlphaFold solved a 50-year-old problem, won the highest prize in science, and generated enormous goodwill — but it doesn't defend Google's position in the model race that determines the company's valuation. Gemini does. In 2026, every large lab is allocating its scarcest resource — elite researchers — toward the same competition, and DeepMind just demonstrated that not even a Nobel-winning team sits outside that calculus.
There's a real scientific bet embedded in the reorganization, and it deserves to be stated fairly: DeepMind believes general-purpose frontier models are becoming good enough at science that one strong model plus automation beats N specialized teams. If that's right, dissolving AlphaFold's structure is rational. If it's wrong, the company traded a proven method for a promising slogan — and the researchers who could rebuild it now work elsewhere. Tellingly, "elsewhere" is often Anthropic, which has been absorbing frontier research talent all year while locking up multi-gigawatt compute to feed it.
The wider pattern
DeepMind isn't acting alone. Across the industry, dedicated science efforts are being absorbed into general model programs, while governments push in the opposite direction — the US Department of Energy's Genesis Mission is explicitly structured around focused, Apollo-style scientific projects. The result is a strange inversion: public programs are adopting the grand-challenge model at the moment its inventor is walking away from it.
For anyone who cheered AlphaFold as proof that AI could serve science rather than just chat, the takeaway is sobering but clarifying. AI-for-science didn't fail on the merits — it lost a budget fight. The next AlphaFold-scale breakthrough will need either a lab insulated from the frontier race, or a frontier model that really can do what a dedicated team once did. We're about to find out which of those exists.
