DeepMind Exodus: Google Loses a Nobel Laureate and the Transformer Co-Author

by Zaira Varghese 29. June 2026 AI 0

The DeepMind exodus has become the defining talent story of the AI race, after Google’s flagship research lab lost a string of marquee scientists to OpenAI and Anthropic inside a single bruising week. The departures rattled investors, erased hundreds of billions in Alphabet market value, and reignited a debate about whether the lab that helped invent modern AI can hold its lead.

What sparked the DeepMind exodus

The unraveling began on Thursday, June 18, when Noam Shazeer — a co-author of the landmark 2017 paper “Attention Is All You Need,” which introduced the Transformer architecture underpinning nearly every large language model — announced he was leaving Google DeepMind for OpenAI. Within 48 hours, John Jumper, who led the Nobel Prize–winning AlphaFold project, said he was departing for Anthropic. For a lab built on its scientific prestige, losing the Transformer co-author and an AlphaFold Nobel laureate in the same week was a gut punch.

The researchers Google lost

The headline names were only part of the story. According to reporting tracking the moves, at least six named researchers have scattered to Meta, OpenAI, and Anthropic in recent months:

Noam Shazeer — Transformer co-author and a Gemini co-lead, to OpenAI.
John Jumper — AlphaFold lead and 2024 Nobel laureate in Chemistry, to Anthropic.
Jonas Adler — a Gemini AI coding lead, to Anthropic.
Alexander Pritzel — a Gemini pretraining specialist and AlphaFold contributor, to Anthropic.

The clustering toward Anthropic is striking. SignalFire’s talent research has noted that DeepMind engineers were roughly 11 times more likely to leave for Anthropic than to move the other way — a flow that predates this week and hints at something structural rather than a one-off poaching spree.

Why the market panicked

Wall Street treated the news as more than gossip. Alphabet shares slid about 7% on June 22, 2026 — the stock’s worst single session in over a year — wiping out roughly $250 billion in market value, according to Bloomberg estimates relayed in Fortune’s reporting on the departures. Investors read the losses as a signal that the talent advantage Google long enjoyed may be eroding just as the competition for frontier models intensifies.

Is the lead really slipping?

Google executives have pushed back, noting that DeepMind still employs thousands of researchers, ships Gemini upgrades at a rapid clip, and retains deep benches in robotics, science, and infrastructure. Star departures, they argue, are an occupational hazard when every lab is paying record packages for a small pool of elite talent.

Skeptics counter that the pattern matters more than any single resignation. When a lab’s most decorated scientists choose rivals — and a Nobel laureate publicly bets on a competitor’s mission — it shapes where the next generation of researchers wants to work. Recruiting compounds: prestige attracts talent, and talent produces the breakthroughs that sustain prestige.

What the DeepMind exodus means next

The immediate question is whether Google can stem the bleeding with retention packages, new mandates, and high-profile launches that remind the field why DeepMind became iconic in the first place. The longer-term story is a reshuffling of AI’s brain trust: OpenAI deepens its architecture expertise, Anthropic stacks scientific firepower around its safety-first pitch, and Meta keeps writing checks. Whatever happens to Gemini’s roadmap, the DeepMind exodus has already redrawn the map of who is building the future of AI — and proven that in this race, people are the scarcest resource of all.