Two Pioneers Depart Google in the Same Week
Google just took a hit it probably saw coming but couldn't stop. Within seven days, two of its most important AI researchers announced they're leaving — one for OpenAI, the other for Anthropic. The timing isn't coincidence. It's a signal about where talent is flowing, and it stings more than usual because both departures involve people who built things the rest of us now take for granted.
Noam Shazeer made his move public on X on June 18. If that name doesn't ring a bell, the paper he co-wrote almost certainly does: "Attention Is All You Need." That 2017 work introduced the Transformer architecture, and yeah — it's the foundation for basically every large language model in use today. Google pulled Shazeer back from Character.AI in 2024 as part of what was reported as a $2.7 billion deal, slotted him in as co-lead on Gemini, and here we are less than two years later. Sam Altman publicly welcomed him at OpenAI.
John Jumper followed with his own announcement. He's been at Google DeepMind for nearly nine years, and he shared the 2024 Nobel Prize in Chemistry for AlphaFold — the system that cracked protein structure prediction, one of those problems biology had been wrestling with for decades. Jumper said he's taking some time off before joining Anthropic, and both sides confirmed the move. His work on protein folding maps pretty cleanly onto Anthropic's expanding push into AI for science.
Two departures. One week. Different labs, different specialties, same underlying story.
What's Going On Inside DeepMind
The departures aren't happening in a vacuum. Bloomberg reported that people inside DeepMind have been raising concerns about something specific: Google doesn't have a clear product for companies building AI coding tools. Anthropic and OpenAI have both gained serious ground in that space, and the internal frustration is real.
It's not just reporters pointing this out. Google's own CEO, Sundar Pichai, said in May that the company was "a bit behind" on agentic coding, and he directly tied that gap to a lack of developer-facing products. That's leadership admitting the problem out loud, which is rarer than you'd think in a company this size.
So now you've got researchers who helped build foundational technology watching Google struggle to ship competitive tools for developers, while the labs they're heading to are actively building in that exact space. It's hard not to see how those two things connect. When the people who understand a technology best start looking elsewhere, it usually means they don't see themselves building the future there anymore.
There's also a quieter dimension here. DeepMind was founded on the idea that AI could solve hard scientific problems, and for a while it actually did. AlphaFold changed the field. But the company's commercial trajectory has drifted toward products and platforms, not the kind of focused tooling that developers actually use day to day. That drift matters when you're trying to convince top talent that their work will ship and matter.
How the Market Read It
Alphabet shares dropped somewhere between 5% and 6% on June 22. Market reports were quick to connect the dots between the departures and broader concerns about AI spending, plus Google's track record of holding onto senior talent. Interestingly, the stock had actually held up in the days right after Shazeer's move was first reported — so it was the combination, not a single departure, that really spooked investors.
Here's why this matters beyond the headline numbers. Shazeer's Transformer work sits under products that the search industry now depends on — AI Overviews, AI Mode, the whole Gemini lineup. Jumper's research demonstrated what AI could actually accomplish in science, and it did so at a level most people thought was still years away. When researchers operating at that tier choose to leave, it changes how investors and competitors read the frontier race. Not because their absence breaks anything today, but because it reshapes expectations about tomorrow.
That said, none of this changes how Gemini, AI Overviews, or AI Mode work right now. The systems keep running. The models keep answering questions. What shifts is the narrative — and narratives matter when you're trying to attract both talent and investment.
What Comes Next
Anthropic has an AI for Science event lined up for June 30, which feels like more than a scheduling coincidence given Jumper's arrival. OpenAI has filed confidentially for an IPO, and both companies have been actively recruiting from larger labs. The pattern is clear: the places building developer tools and science-focused AI are where top researchers want to be.
The open question for Google is retention. The company paid a massive price — reports say $2.7 billion — to bring Shazeer back once, and it couldn't keep him. That's an expensive lesson. Whether Google changes how it holds onto senior researchers, or whether this becomes a pattern we see repeated, is what to watch over the next few quarters.
The technology itself isn't going anywhere. But the people who helped build it are moving, and that tells you something about where the industry is heading.