Google Dismantles AlphaFold Team as AI Strategy Pivots to Gemini
Google DeepMind has disbanded its award-winning AlphaFold team, reassigning key members as the company refocuses on Gemini AI agents. The shift marks a…
Google has dismantled the core team behind AlphaFold, one of artificial intelligence's most celebrated breakthroughs, according to Financial Times. Most key members and original paper authors have been reassigned to other projects, while several have left the company entirely.
AlphaFold, which solved a 50-year-old puzzle in molecular biology, can predict three-dimensional protein structures from amino acid sequences in minutes—work that previously took researchers years. The program now underpins drug discovery, vaccine development, and research into neurodegenerative diseases like Alzheimer's and Parkinson's.
Google DeepMind launched the project in 2018. By 2020, the team's breakthrough was recognized as the solution to the "protein folding problem," the fundamental question of how amino acids fold into the complex shapes that determine a protein's biological function. Scientists had identified only roughly 170,000 protein structures over five decades using X-ray and nuclear magnetic resonance techniques before AlphaFold accelerated the process dramatically.
In 2021, Nature published the team's methodology alongside structure predictions for the entire human proteome—the complete catalog of proteins our species produces. DeepMind then released the AlphaFold Protein Structure Database, providing free public access to over 200 million protein structure predictions. The impact was immediate and global: researchers worldwide began using it to understand disease mechanisms and design new treatments.
The recognition came at the highest level in 2024. Demis Hassabis, DeepMind's CEO, and John Jumper, who rose from staff research scientist to VP and engineering fellow, shared the Nobel Prize in Chemistry for the work. Jumper announced in June he was leaving DeepMind to join Anthropic, taking several colleagues with him.
The team's dissolution reflects a broader strategic reorientation at Google. The company is consolidating resources behind Gemini, its answer to OpenAI's ChatGPT and other large language models. Most visibly, Google has begun rolling out Gemini Spark, an AI agent designed to automate workflows across its Workspace ecosystem—Gmail, Docs, Sheets, and Calendar. The service launched in the United States in May and expanded to Hong Kong this week, powered by the Gemini 3.6 Flash model.
Gemini Spark handles tasks like summarizing email threads and logging customer feedback into spreadsheets automatically. Unlike open-source AI tools that require custom integrations, Spark operates natively within Google's infrastructure, keeping user data within the company's secure cloud environment and running continuously even when devices are offline.
"Google's deep integration across Workspace tools is certainly an advantage for everyday productivity," said Hu Qi, a doctoral researcher specializing in AI agent security at the University of Hong Kong. "However, when evaluating the raw reasoning capabilities of the underlying language model, many users in academic and technical communities still favour models like ChatGPT for specialised tasks."
The shift underscores how rapidly AI investment priorities are changing. While AlphaFold represented pure research—solving a scientific puzzle with applications that unfolded over time—Gemini agents target immediate, measurable productivity gains in consumer and enterprise markets. Google faces intense competition from OpenAI, Anthropic, and open-source models, all vying for dominance in the generative AI space. DeepMind's restructuring suggests Google believes its future competitive advantage lies not in specialized scientific breakthroughs but in seamlessly embedding AI into the tools billions of people already use daily.
