Google loses four top AI researchers to Anthropic and OpenAI in a single week as talent competition intensifies

Four of Google's most prominent AI researchers departed in a single week for rival labs, with three heading to Anthropic and one to OpenAI. The exits — including a Nobel laureate and one of the architects of the modern transformer — mark a notable acceleration in the talent competition between Google and its rivals as Anthropic and OpenAI each move toward public offerings.

What's new

The confirmed departures as of June 24, 2026:

• John Jumper — former Google DeepMind director and 2024 Nobel Prize winner in Chemistry for his work on AlphaFold. Joining Anthropic.
• Jonas Adler — key contributor to Google's AI coding efforts and Gemini model development. Joining Anthropic.
• Alexander Pritzel — key contributor to Google's AI training processes and Gemini model. Joining Anthropic.
• Noam Shazeer — AI researcher at Google since 2000, widely known for co-authoring the "Attention Is All You Need" paper that introduced the transformer architecture. Joining OpenAI.

Bloomberg first reported the Adler and Pritzel departures, citing people familiar with the matter. At least one departure was preceded by a computing resource dispute, with resources reassigned from the researcher's project to other initiatives.

Context

The exits follow a period of accelerating talent pressure on Google. Noam Shazeer is among the most historically significant researchers in the field — he co-wrote the 2017 paper that introduced transformers and spent two decades at Google before leaving to found Character.AI; his decision to join OpenAI now represents a further shift in the competitive balance. John Jumper's AlphaFold work was central to Google DeepMind's scientific reputation.

Anthropichas been particularly aggressive in drawing Google talent. Three of the four departures are headed there directly. Google's competitive position in frontier AI has been under scrutiny in 2026 following delays to Gemini 3.5 Pro and the broader context of OpenAI and Anthropic releasing increasingly capable models.

Why it matters

Individual researcher moves rarely move the needle on their own — AI products are built by large teams, and Google continues to employ thousands of researchers and engineers. But the pattern here is notable: Shazeer and Jumper are not ordinary researchers, and losing both of them in the same week as two key Gemini contributors suggests something structural, not just opportunistic.

The practical mechanism is equity. As both Anthropic and OpenAI prepare for public offerings, the option-value gap between working at an early-stage private company and working at one of the world's largest public companies has widened sharply. Google can offer compensation competitive with current salaries, but cannot match the pre-IPO equity upside of a company valued in the hundreds of billions that hasn't yet listed.

For the broader AI ecosystem, these moves signal where researchers themselves believe the frontier is moving — and which organizations they want to be part of when it gets there.