Jonas Adler and Alexander Pritzel have become the latest senior researchers to depart Google for Anthropic, according to a June 24 Bloomberg report. Both individuals are identified as key contributors to Google’s Gemini model, with Adler focusing on AI coding efforts and Pritzel specializing in pretraining (the initial phase where models learn from large data sets). Neither company has formally confirmed the move, and the sources of the information remain anonymous.

These departures follow a period of significant talent attrition at Google DeepMind. Last week, Noam Shazeer, a vice-president of engineering and co-author of the 2017 Transformer architecture paper, left for OpenAI. Shortly thereafter, Nobel laureate John Jumper, known for his work on AlphaFold, also announced his exit for Anthropic. Jumper's resignation led to a decline in Alphabet’s share price of approximately 6%, representing over $245bn in market value during a single session.

Industry analysts attribute this migration to several factors, primarily the potential financial upside of pre-IPO equity at companies like Anthropic and OpenAI. While Google maintains immense market capitalization, researchers increasingly value the equity potential offered by fast-growing private labs compared to established corporate salaries. Additionally, internal competition for computational resources has influenced these decisions. External demand for Google’s custom chips, known as Tensor Processing Units (TPU), has created queues that restrict internal research access, causing some experts to seek more consistent hardware availability elsewhere. Anthropic, notably, is among the external customers purchasing this capacity from Google.

Jonas Adler and Alexander Pritzel have become the latest senior researchers to depart Google for Anthropic, according to a June 24 Bloomberg report. Both individuals are identified as key contributors to Google’s Gemini model, with Adler focusing on AI coding efforts and Pritzel specializing in pretraining (the initial phase where models learn from large data sets). Neither company has formally confirmed the move, and the sources of the information remain anonymous.

These departures follow a period of significant talent attrition at Google DeepMind. Last week, Noam Shazeer, a vice-president of engineering and co-author of the 2017 Transformer architecture paper, left for OpenAI. Shortly thereafter, Nobel laureate John Jumper, known for his work on AlphaFold, also announced his exit for Anthropic. Jumper's resignation led to a decline in Alphabet’s share price of approximately 6%, representing over $245bn in market value during a single session.

Industry analysts attribute this migration to several factors, primarily the potential financial upside of pre-IPO equity at companies like Anthropic and OpenAI. While Google maintains immense market capitalization, researchers increasingly value the equity potential offered by fast-growing private labs compared to established corporate salaries. Additionally, internal competition for computational resources has influenced these decisions. External demand for Google’s custom chips, known as Tensor Processing Units (TPU), has created queues that restrict internal research access, causing some experts to seek more consistent hardware availability elsewhere. Anthropic, notably, is among the external customers purchasing this capacity from Google.