Four Google AI Leaders Gone in Six Days — Gemini Co-Lead and a Nobel Laureate Head to Anthropic and OpenAI

Bloomberg reported on June 24 that Jonas Adler and Alexander Pritzel, both viewed internally as key contributors to Google's Gemini AI model, are set to move to Anthropic. Adler worked on the company's AI coding effort, while Pritzel focused on pretraining — the early stage in which a model learns from large volumes of data. They are the second pair to head for Anthropic within a week, adding to a recent run of Google departures.

Four in Six Days: Who Left

These are not isolated exits but a tightly clustered wave. Between June 18 and 24, researchers anchoring Google's AI front line left, one after another, for rivals.

Two of the names carry outsized weight. Noam Shazeer was a co-author of the landmark 2017 paper "Attention Is All You Need," which introduced the transformer architecture. Google reportedly paid around $2.7 billion in August 2024 to bring him back from his own startup, Character.AI — yet less than two years later he is heading to OpenAI. In a post on X, Shazeer wrote that he was "excited to share that I'll be joining OpenAI and look forward to working with the exceptional team there."

John Jumper is a co-creator of AlphaFold, the protein-structure-prediction AI, and shared the 2024 Nobel Prize in Chemistry with Google DeepMind's Demis Hassabis. That a researcher operating at the very summit of basic science would move to a rival underscores the significance of this wave.

Why Now — The Pull of a Pre-IPO Payday

The one-directional flow has a clear logic. Anthropic and OpenAI are both seen as close to going public, offering even well-compensated Big Tech employees a chance at a rare payday by signing on before an IPO and receiving equity ahead of a listing.

The two companies' steep growth adds to the case. Anthropic has said its run-rate revenue surpassed $30 billion in 2026, up from roughly $9 billion at the end of 2025, and that the number of business customers spending more than $1 million a year doubled to over 1,000 in less than two months. To support that surging demand, Anthropic expanded its partnership with Google and Broadcom, securing access to roughly 3.5 gigawatts (GW) of Google TPU compute via Broadcom starting in 2027.

The Cost to Google

On the surface, Google still commands enormous capital, infrastructure, and a deep research bench. But losing the people responsible for the foundations of model competitiveness — Gemini's coding and pretraining work — all at once is not trivial. Pretraining and coding are central axes that directly shape the performance of today's AI models.

What It Means

This cascade signals that AI competition has decisively shifted beyond a contest of model performance into a war for talent. With Anthropic and OpenAI riding capital-market expectations as they near IPOs, expect aggressive talent absorption to continue for some time. The key thing to watch is which levers — compensation, research autonomy, speed to product — Google uses to stem the flow.