On June 19, John Jumper posted the kind of message that ends an era. After nearly nine years at Google DeepMind, the man who co-built AlphaFold and won the 2024 Nobel Prize in Chemistry for it said he was leaving. His destination: Anthropic.

He wasn't the first that week. He wouldn't be the last.

The day before, Noam Shazeer announced he was walking out of Google to join OpenAI. Shazeer isn't just any engineer. He is one of the eight co-authors of "Attention Is All You Need," the 2017 paper that introduced the Transformer — the architecture underneath ChatGPT, Gemini, Claude, and basically every large language model you have ever touched. Google published that paper. Google is now watching its authors leave for the companies that turned it into a business.

And Shazeer's exit stings in a particular way. He left Google in 2021 to co-found the chatbot startup Character.AI. In 2024, Google paid roughly $2.7 billion to license that startup's technology and bring Shazeer back into the fold. Barely eighteen months later he's gone again — this time to the one company Google fears most. That is an expensive round trip.

Then came the aftershock. On June 24, Bloomberg reported that two more DeepMind researchers, Jonas Adler and Alexander Pritzel, were also headed to Anthropic. Adler worked on Google's AI coding tools. Pritzel worked on pretraining, the expensive foundational stage where a model learns from oceans of data. Both had contributed to AlphaFold alongside Jumper. Four departures. One week. Two rivals.

Wall Street noticed. Alphabet shares dropped roughly 5% around June 22, and analysts openly tied the slide to fears that Google can't hold onto the people who make its AI worth anything.

Here is the part nobody at Google wants to say out loud: this isn't really about salary. Alphabet can outbid almost anyone on cash. What it cannot manufacture is a pre-IPO cap table.

Anthropic has filed to go public at a reported $965 billion valuation. OpenAI is lining up its own IPO with Goldman Sachs and Morgan Stanley. The math is brutal in its simplicity. A fraction of a percent of a $965 billion company that hasn't listed yet can dwarf a decade of Google salary. Equity like that exists once — in the window before the IPO — and that window is closing fast at both labs. A researcher choosing between them isn't choosing between paychecks. They're choosing between a paycheck and a fortune.

My Opinion

I'll be blunt: Google is getting exactly what it built. For a decade it treated research like a trophy case. Publish the landmark paper, collect the Nobel, let the org chart swell, and assume the talent would stay because the campus is comfortable. Then Anthropic and OpenAI showed up and did the unglamorous thing. They shipped products developers actually pay for.

DeepMind staff have reportedly grumbled that the lab still lacks a clear enterprise product in AI coding — the exact arena where Claude and GPT are printing money. That's the real wound. Smart people don't leave because they're underpaid. They leave because they want their work to ship, and they have watched their own inventions become someone else's revenue.

Google will spin this as normal churn. It isn't. When the co-author of the paper that created your entire industry and a Nobel laureate both walk out the door in 48 hours, that's not turnover — that's a verdict. The next twelve months will decide whether Gemini can survive losing the people who taught it to think, or whether Google spends the AI decade the way it spent the mobile one: inventing the future and letting other people cash it in.

Author: Yahor Kamarou (Mark) / www.humai.blog / 03 Jul 2026