Alphabet Drops 6% After Losing Shazeer to OpenAI and Jumper, AlphaFold Nobel Laureate, to Anthropic
The departure of Gemini co-leader to OpenAI and John Jumper to Anthropic led Alphabet to its worst daily decline in 13 months. Google paid $2.7 billion for Shazeer less than two years ago.
Alphabet closed down about 6% on Monday, marking its worst session since May 2025, following the loss of two key figures in less than four days that reopened the debate on who controls the frontier of artificial intelligence research. On the prior Wednesday, June 18, Noam Shazeer, Vice President of Engineering and co-leader of the Gemini model, announced his departure to OpenAI. Then on Friday, John Jumper, head of AlphaFold at Google DeepMind and winner of the 2024 Nobel Prize in Chemistry, communicated his move to Anthropic after nearly nine years in the division.
The Cost of Departure
Shazeer's departure carries a double weight. He co-authored the 2017 paper "Attention Is All You Need," which introduced the Transformer architecture, the foundation of nearly every language model in production today, including Gemini itself. He is the same individual that Google brought back from Character.AI less than two years ago for about $2.7 billion in a talent acquisition and licensing deal. Therefore, Alphabet effectively pays twice for the same signature: once for the buyback and once for the departure.
The loss of Jumper touches on another wound. AlphaFold has cataloged over 200 million protein structures and has become a pillar of future revenue at DeepMind, with licensing agreements for pharmaceutical companies. In his message on platform X, Jumper thanked Demis Hassabis for "taking a real risk" by placing him in charge of the project six months after his PhD. The immediate reading on Wall Street is that DeepMind's healthcare pipeline loses execution momentum, and Anthropic gains a name whose scientific authority in life sciences is worth more than any press release hire.
CapEx and Distrust
Analysts approached Alphabet with two questions. The first is whether the pace of CapEx in data centers still makes sense if the team operating the frontier could become fragmented. The second is whether Anthropic and OpenAI have managed, for the first time, to assemble a research proposal that is more attractive than DeepMind's. Anthropic has been allocating 500,000 Trainium2 chips to AWS's Project Rainier, with plans for scaling up to 1 million chips, while holding $30 billion in Azure commitments and maintaining access to Google Cloud's TPUs. For a scientist needing to run protein simulations at scale, this type of computational package is far more compelling than an accelerated vesting schedule in stock.
Read-Across for London, Seoul, and São Paulo
The immediate effect does not stay in Mountain View. DeepMind has operations concentrated in London, and the reorganization of priorities around AlphaFold is likely to redraw contracts with British hospitals and NHS partners that had been banking on a roadmap centered on Jumper's leadership. In South Korea, where Anthropic opened an office in Seoul on June 17 focusing on partnerships with Samsung and SK Hynix, the arrival of a Nobel laureate strengthens the sales thesis in biomedical research, a segment that OpenAI has yet to cover with the same depth. In India, the main offshoring hub for Big Tech, the market signal is the same: companies with a senior talent-dependent AI model now compete for a fast-tracked talent transfer population. In Brazil, where Itaú, Bradesco, and Petrobras operate multi-year cloud contracts with Google in the hundreds of millions of dollars range, any delay in the release cadence of Gemini is likely to open a window for Microsoft Azure and AWS to reposition competitive proposals during renewal cycles.
What Doesn’t Change
For Anthropic, hiring Jumper accelerates the path to academic validation in biological sciences, an area where Mythos 5 still competes under conditions of inferiority against Gemini. For OpenAI, reintroducing Shazeer into the Frontier Alliances roadmap reopens the race for the next capability paper. For Alphabet investors, the less appealing part of the reorganization remains: paying dearly for the same scientific signature twice and still not knowing how long the adaptation period will be until the next generation of Gemini without Shazeer at the technical helm.