# Google DeepMind disbanded its Nobel Prize-winning AlphaFold team and the talent went straight to Anthropic > Source: > Published: 2026-07-29 11:19:09+00:00 *Google DeepMind has broken up the AlphaFold group that gave it its cleanest scientific win. The timing looks worse because John Jumper and other AlphaFold veterans are now headed toward Anthropic.* There's a particular kind of institutional confidence required to take apart the team behind your greatest scientific achievement, and Google DeepMind has just shown it. The Financial Times reported this week that DeepMind has disbanded the AlphaFold team, with researchers reassigned to Gemini-based projects, enzyme design, nuclear fusion, genomics and Alphabet's drug-discovery company Isomorphic Labs. About a quarter have left the company, according to the same report. The team no longer exists. That matters because AlphaFold was not a side project that happened to get flattering headlines. Demis Hassabis and John Jumper shared half of the 2024 Nobel Prize in Chemistry for protein structure prediction, while David Baker won the other half for computational protein design, according to the Nobel committee. AlphaFold was the DeepMind example you could point to when someone asked whether modern AI could do more than write passable emails and summarise meetings. It could. The departures make the shift harder to brush off. In June 2026, Jumper left Google DeepMind for Anthropic after nearly nine years. TechCrunch reported that he announced the move in a post on X, where he credited Hassabis with letting him lead AlphaFold only six months after finishing his PhD. Bloomberg then reported that Jonas Adler and Alexander Pritzel, both AlphaFold authors and Gemini contributors, were set to move to Anthropic too. That's three serious names from the AlphaFold orbit moving toward the same rival in a matter of days. Gone in days. The same stretch brought another ugly headline for Google. Reuters reported on June 17 that Noam Shazeer, a Google vice president of engineering, Gemini co-lead and co-author of the 2017 Transformer paper, said he was leaving for OpenAI. Alphabet shares then fell around 5% on June 22, with Bloomberg tying the sell-off to the Jumper and Shazeer exits as well as wider weakness in large tech stocks. The market reaction was blunt, even if markets are not always wise. ## AlphaFold was already real science AlphaFold's record is hard to overstate without sounding as if you're writing a company brochure, so stick to the numbers. DeepMind's own AlphaFold page says the system has predicted more than 200 million protein structures and has more than three million users in over 190 countries. The Nobel committee said AlphaFold2 had helped researchers understand antibiotic resistance and create images of enzymes that can decompose plastic. That is not a demo. It is working scientific infrastructure. The model isn't disappearing. Isomorphic Labs still exists, and its own site says it is building on and beyond AlphaFold for drug discovery. DeepMind and Isomorphic also published a joint bioresilience approach on July 16, 2026, showing that Alphabet still wants the science story attached to the company. But keeping the brand and keeping the team are not the same thing. That detail matters. Research teams carry habits, arguments and half-formed ideas that don't survive neatly in a memo. You can reassign people to Gemini, fold some into Isomorphic and scatter others across nuclear fusion or genomics, but you shouldn't pretend the next AlphaFold is guaranteed to emerge from the same machinery. The system may keep improving. The original engine has been taken apart. ## Google is choosing Gemini first The answer is Gemini. DeepMind's new direction, as described by the Financial Times, is less about one dedicated group solving one defined scientific problem and more about building Gemini-powered systems that can assist scientists and eventually automate parts of research. That isn't automatically wrong. A general model that helps across biology, chemistry and physics could end up doing more than one specialist group ever could. But it is different. Frankly, the timing is awkward for Google. DeepMind built enormous credibility on the idea that beneficial AI would show up in concrete scientific breakthroughs, not just bigger chat windows. AlphaFold was the proof point. Disbanding the team behind it in the same period that Jumper leaves for Anthropic does not look like calm succession planning from the outside. It looks like the Gemini land grab consuming everything within reach. Anthropic now gets to tell a different story. It already has Claude in the frontier model race, and now it is adding a Nobel laureate whose best-known work turned machine learning into a tool that millions of researchers actually use. Adler and Pritzel are not household names, but their author lists put them inside the work that made AlphaFold matter. If you're trying to build AI systems that scientists trust, that kind of talent is not ornamental. Google may still be right. Gemini-powered scientific automation could beat the old model of dedicated teams working on bounded problems one at a time. But that case has to be made with results, not org charts. For now, Anthropic has the AlphaFold names, and DeepMind has the harder job: proving that breaking up its most celebrated science team was not a mistake. **Also read:** [OpenAI's own AI agents broke out and hacked Hugging Face and a second company](https://startupfortune.com/openais-own-ai-agents-broke-out-and-hacked-hugging-face-and-a-second-company/) • [Dwelly raises $170M to buy up UK letting agencies one by one and run them on AI](https://startupfortune.com/dwelly-raises-170m-to-buy-up-uk-letting-agencies-one-by-one-and-run-them-on-ai/) • [Moonshot AI closes round at $31.5 billion and is already chasing $50 billion](https://startupfortune.com/moonshot-ai-closes-round-at-315-billion-and-is-already-chasing-50-billion/)