The Geography and Economics of AI Talent Migration

Four senior AI researchers just walked out of Google in a few weeks. It wasn't about the paycheck. Not seniority. Not even their specific roles.

It was London. That’s what ties all this mess together.

What’s actually happening? People are leaving. Jonas Adler and Alexander Pritzel , both key players in that Gemini model stuff. Adler worked on coding AI, pushing those code efforts. Pritzel sat deep inside the training pipeline the part that decides how models even behave. Both are heading over to Anthropic now.

They follow a trail. John Jumper , the Nobel guy who co-led AlphaFold. Then Noam Shazeer , who wrote one of the original transformer papers back in 2017. He’d been at Google for more than twenty years before jumping ship to OpenAI. A long way from here.

Google still claims they’re fine with where they stand in this AI talent market. But look at what's actually moving.

Why London is the thread. It comes down to where the real heavy lifting seems to be happening. DeepMind’s core research leadership, it sits right there in London. And those researchers are stuck by these non-compete aGreements. The kind UK law actually enforces, unlike California, where you can just shrug those clauses off. That legal difference shows up immediately in how these exits are being structured.

Jumper, for instance. People familiar with his plans say he won't start at Anthropic until next year. A delay. It looks less like a simple break and more like some kind of contractual cooling-off period built around those non-competes. Just waiting out the clock.

The pattern isn't just about the big names, though. Arthur Conmy announced he’s joining Anthropic to tackle model alignment. He said plainly he’s moving from London over to San Francisco for that job. It makes sense.

Researcher Lucas Beyer saw this coming early on. He flagged it. These departures? Mostly long-time Londoners. And he linked it to something structural. The center of gravity for all the actual AI pre-training seems to be shifting toward Mountain View. If the most important research work is migrating to California, why wouldn't people follow that pull? Non-competes just make sense then.

There’s internal evidence too. Something about compute. Shortly before Shazeer left, computing power assigned to one of his projects got reportedly shunted over to a London DeepMind team. They needed better collaboration on pre-training. Compute is the currency in this whole field. Watching that resource get reassigned? That tells you where the real focus is shifting. It’s how researchers read the room when things are changing underneath.

Why now specifically? Geography explains who leaves. Timing explains why it's happening right now. Anthropic and OpenAI are both circling public listings. And a pre-IPO offer has something that a standard Google salary just can't match. Equity. That could be the defining payday the moment either company decides to go public.

Anthropic’s numbers make this pull especially hard to ignore. They recently raised funding, hitting a valuation near nine hundred sixty-five billion dollars. That already puts them ahead of OpenAI. And they are reportedly aiming for a public listing as soon as this fall hits. SignalFire did some analysis back in 2025. It found that DeepMind engineers were roughly eleven times more likely to leave for Anthropic than the reverse. A ratio that turns scattered individual exits into something almost steady.

Investors noticed too. Alphabet shares took a hit. Seventy-two percent drop intraday after Jumper exited. That was the steepest single-day plunge since February. A stock reacting that violently to one researcher leaving tells you how seriously the market is taking this whole talent war unfolding. It’s not just academic; it's financial panic mixed in with the tech shift.

Demis Hassabis, the head of DeepMind, pushed back at an event in Cannes. He called the current AI talent market the most ferociously competitive thing the tech industry has ever seen. He insisted Google still gets its fair share of top researchers. He’s not wrong about the intensity. But four senior departures in two weeks? Most traced right back to that same city. And those shifting compute resources? It suggests this isn't just fierce competition anymore. It feels geographic. The map is pointing away from London now.

It’s a mess of movement, really. People chasing opportunity wherever the legal structure allows it. Compute moves where the money flows. And the geography dictates the permission to move. Everything seems tied together by that subtle, inescapable shift in gravity.