Google’s Gemini team is running into an awkward problem: the people who helped build it keep leaving. According to Bloomberg, Jonas Adler and Alexander Pritzel are moving to Anthropic, adding to a string of exits that has already included Noam Shazeer and John Jumper. For a company that once looked untouchable in frontier AI, that is a lot of talent to lose in a very short span.
The timing is especially rough because these are not random hires. Adler and Pritzel were part of the group behind Gemini’s architecture and training, which makes them more than just familiar names on an org chart. When the engineers who know how the machine was assembled start heading for the door, rivals do not need to guess where the weak spots are.
Who Google has lost from its AI ranks
Shazeer’s departure already carried symbolic weight. He had been at Google since 2000, aside from a few years at Character.AI, and Google later paid about $2.7 billion to bring Character.AI back into its orbit largely so he could keep working on Gemini. Jumper’s exit stung in a different way: as DeepMind research director and a 2024 Nobel Prize winner in chemistry, he was one of the most visible scientific figures in the company.
That is the context behind the latest pair of departures. Google is not just losing employees; it is losing credibility in the one area where it was supposed to have an edge: foundational AI research. Competitors like OpenAI and Anthropic have made the talent race as important as the model race, and they are increasingly willing to pay for both.
Why Anthropic keeps looking attractive
Anthropic and OpenAI are both widely viewed as potential public-market candidates, which gives them a new weapon in recruitment: equity with a possible upside story. That matters in a sector where cash alone is no longer enough to pry away elite researchers. If an IPO is even on the horizon, stock grants suddenly look a lot more exciting than a safe, predictable corporate package.
- Google has now lost several high-profile AI figures tied to Gemini and DeepMind.
- Anthropic and OpenAI can sell researchers on salary, equity, and a potential listing bonus.
- The real fight is no longer just about compute and model quality, but about recruiting scarce frontier-AI talent.
What Google’s Gemini team losses mean for the future
Google still has scale, cash, and deep research infrastructure, so no one is declaring the company out of the race. But repeated departures make it harder to argue that Gemini’s core team has the same stability it did a year ago. The next question is whether Google responds with a louder retention push, a broader management shake-up, or simply more resignations dressed up as normal industry churn.