First it was the Transformer co-inventor. Then a Nobel laureate. Now two more Gemini contributors. Google is watching its AI bench walk out the door – and the pace is accelerating.
On Tuesday, Bloomberg reported that Jonas Adler and Alexander Pritzel – both viewed internally as key contributors to Google’s Gemini model – are leaving for Anthropic.
Adler worked on Google’s AI coding efforts. Pritzel was involved in pretraining, the foundational phase where models learn from massive datasets.
Both also contributed to the AlphaFold protein-folding research that won a Nobel Prize.
That brings the tally to four senior AI departures in six days. Three went to Anthropic. One to OpenAI. All from teams working on Google’s most important AI projects.
The Full Scorecard
Here’s how the last week unfolded for Google:
June 18 – Noam Shazeer announces he’s leaving for OpenAI.
He co-authored the 2017 “Attention Is All You Need” paper that invented the Transformer architecture. Google paid $2.7 billion to bring him back from Character.AI less than two years ago. He was VP of Engineering and co-lead of Gemini.
June 20 – John Jumper announces he’s leaving for Anthropic. He won the 2024 Nobel Prize in Chemistry alongside DeepMind CEO Demis Hassabis for AlphaFold. Nine years at DeepMind. Gone.
June 24 – Jonas Adler and Alexander Pritzel reportedly heading to Anthropic. Both worked on Gemini. Both worked with Jumper on AlphaFold.
Three Gemini contributors heading to Anthropic in the same week isn’t coincidence. It looks like targeted recruitment of people who know exactly how Google’s flagship model works.
The IPO Magnet
The timing isn’t mysterious. Both Anthropic and OpenAI are preparing for what could be among the largest IPOs in tech history. Anthropic recently raised $65 billion at a $965 billion valuation.
OpenAI filed confidential IPO paperwork earlier this month.
Pre-IPO equity is the magnet. Google can match salary. It can’t match the upside of joining a company worth nearly a trillion dollars before it goes public.
For researchers sitting on Google RSUs that vest predictably, the math changes when a startup offers shares that could multiply overnight.
It’s Not Just Money
Bloomberg uncovered something that suggests the problem runs deeper than comp. Shortly before Shazeer announced his move to OpenAI, Google reassigned computing power from one of his projects to a London-based DeepMind team.
The move was meant to boost collaboration and streamline pretraining work. But for someone like Shazeer – who left Google once before because he thought it was too cautious about shipping – having resources pulled from his project may have been the final straw.
Meanwhile, key members of Jumper’s AlphaFold team have also exited Google DeepMind in recent months. Some went to Isomorphic Labs, an Alphabet spinout focused on AI-designed drugs. Others, clearly, went to Anthropic.
According to a 2025 analysis by SignalFire, DeepMind engineers were nearly 11 times more likely to leave for Anthropic than the reverse.
Google’s Response Looks Weak
Google launched a $450,000 incubator program for departing employees – nicknamed the “Xoogler” fund – just yesterday.
Today, two more people left. As FourWeekMBA noted: “The incubator funds people after they leave. It doesn’t prevent them from leaving.”
Google’s Gemini 3.5 Pro has also reportedly been delayed until July for final adjustments. Losing this many senior researchers right before a major model launch doesn’t help the narrative.
Jefferies, for its part, has maintained a positive investment outlook on Alphabet. But investors are watching.
Google pioneered most of the technology that powers modern AI. The Transformer was invented there. AlphaFold was built there. The question now is whether the company can keep building at the frontier when the people who know how are walking across the street.
Anthropic is building its next model with Google’s best researchers. That sentence would have been unthinkable two years ago. It’s just a fact now
