DeepMind loses Nobel Prize winner to Anthropic

What happened

Noam Shazeer, who helped build Google's LaMDA chatbot, announced he is joining OpenAI after returning to Google in 2024 via its $2.7 billion（約4300億円） licensing deal for Character.ai. Days later, John Jumper, who shared the 2024 Nobel Prize in chemistry for creating AlphaFold, said he is joining Anthropic. The departures sent Google's shares tumbling more than 5% on Monday.

Why it matters

Google DeepMind's top models—Gemini 3.5 Flash and Gemini 3.1 Pro—are often ranked outside the top five places on AI benchmark leaderboards and have fallen behind models from Anthropic, OpenAI, and Chinese labs such as Zhipu AI and MiniMax. The lab's pace of model development also appears slower: Gemini 3.5 Pro is coming out about four months after its last frontier wide release (Gemini 3.1 Pro in February), whereas Anthropic released two Claude Opus updates and debuted a new class of models called Mythos in that same period. Current and former employees describe Google DeepMind's culture as bureaucratic and risk-averse, echoing criticisms Shazeer made publicly years ago.

What to watch

Jumper's move to Anthropic appears connected to the company's stated intention to do more work in biology and science applications—an area where Google DeepMind's focus seems to have diminished relative to its pre-ChatGPT priorities. Both OpenAI and Anthropic are expected to IPO in the coming months, which may be a factor in attracting senior talent seeking equity upside.