Google AI researchers Jonas Adler and Alexander Pritzel set to join Anthropic

Google’s AI bench took another hit as Jonas Adler and Alexander Pritzel prepared to leave for Anthropic, adding two more researchers from Gemini’s orbit to a steady flow of departures out of Alphabet’s AI operation. Adler worked on Google’s AI coding efforts and Pritzel on training systems, and both were viewed internally as important contributors to Gemini.

The moves followed John Jumper’s June 19 decision to leave Google DeepMind for Anthropic and earlier loss of Noam Shazeer, showing that the competition for AI talent has moved beyond startup poaching and into the core teams inside the biggest labs. Google DeepMind’s own history makes the exits more striking: the company says Jumper and Demis Hassabis won the 2024 Nobel Prize in Chemistry for AlphaFold.

The pressure inside Google is not only about prestige. Some employees have grown frustrated over how scarce computing power is allocated, and in at least one case resources for a project were reassigned to another DeepMind team. That kind of internal scramble matters in a field where researchers want freedom to run experiments and enough compute to turn them into products.

Anthropic has the capital and infrastructure story to match the talent push. On May 28, the San Francisco startup said it raised $65 billion in Series H funding at a $965 billion post-money valuation. On April 6, it disclosed a multigigawatt TPU capacity agreement with Google and Broadcom that is expected to begin coming online in 2027.

Google is still trying to show it can keep pace. At its May 19 I/O keynote, the company said it was moving from assistive AI to agents that can independently navigate complex tasks, while unveiling Gemini 3.5 and upgrades to its Antigravity platform. The departures suggest the fight is now being waged on two fronts: model quality and whether top researchers believe the best freedom, compute and commercial upside sit at Google, Anthropic or somewhere else.