DeepMind Abandons AlphaFold Team, Shifts Focus to General AI and Gemini

A major shake-up is underway at Google’s elite AI lab DeepMind. The institution behind the Nobel Prize-winning system that cracked the code on predicting three-dimensional protein structures has quietly dismantled its star team AlphaFold. This restructuring marks a significant departure from DeepMind’s previous strategy, which centered around conquering single grand scientific challenges.

The move comes as part of a broader shift in focus for DeepMind. The company is now fully committed to building general artificial intelligence and agents around its Gemini large language model. According to sources familiar with the matter, most authors of the original AlphaFold paper have been reassigned to other teams over the past year. Among those still at DeepMind, nearly a quarter have left the company entirely.

Those who remain are being integrated into projects built around Google’s Gemini or new directions such as enzyme design, nuclear fusion, and genomics. Another portion has been transferred to Alphabet’s drug discovery subsidiary, Isomorphic Labs. This restructuring is part of a fundamental change in DeepMind’s approach, according to Research Vice President Pushmeet Kohli.

Kohli had previously formulated a nine-year strategy focused on grand scientific challenges. However, he now indicates that this approach has undergone a significant shift. DeepMind is no longer deploying its forces solely around single scientific problems like protein folding but is instead shifting its focus to building Gemini-powered systems that can assist scientists in conducting research and ultimately automate parts of the scientific research process.

The most striking aspect of this restructuring is the talent exodus at the core level. Research engineer John Jumper, who won the Nobel Prize in Chemistry for his contributions to AlphaFold, was transferred earlier this year alongside fellow researcher Jonas Adler to a newly formed internal Google team called ‘Code Strike.’ The mission of Code Strike is to enhance the company’s AI coding capabilities and catch up with rivals Anthropic and OpenAI.

However, this move failed to retain top talent. Last month, Jumper announced his departure for Anthropic, followed by Adler and another AlphaFold colleague, Alexander Pritzel. An anonymous DeepMind employee described the trio as ‘key, important, core members,’ noting that their collective departure caused considerable shockwaves within the company.

The AlphaFold story began in 2018 with its breakthrough use of artificial intelligence to predict three-dimensional structures of proteins. This achievement fundamentally rewrote the path of biological research and paved the way for the creation of Isomorphic Labs. The system’s success ultimately sent DeepMind CEO Demis Hassabis and Jumper to the podium to receive the 2024 Nobel Prize in Chemistry.

When questioned about the team’s dissolution and talent drain, DeepMind expressed pride in its scientific heritage and AlphaFold’s global impact. However, between the lines, it is clear that a legacy built on independent problem-solving has been transitioned into feeding general-purpose large models and agents.