The brain drain from Google to OpenAI and Anthropic shows that in AI money is no longer enough to retain talent

In barely seven days, Google has seen the departure of names that are hard to replace. According to Bloomberg, reported by TechCrunch, researchers Jonas Adler and Alexander Pritzel, both tied to the development of Gemini, are joining Anthropic. Before them, Noam Shazeer —at the company since 2000, with a stint at Character.AI— had left for OpenAI, and John Jumper, co-awarded the 2024 Nobel Prize in Chemistry for AlphaFold, also headed to Anthropic. That an active Nobel laureate changes organizations is extraordinary; that he does so toward the most direct rival adds a symbolic weight that transcends the individual.

The most telling detail is economic. Google went so far as to acquire Character.AI for around 2.7 billion dollars, an operation whose partial motivation was, according to the article, to bring Shazeer back to work on Gemini. That this outlay was not enough to retain him sums up the problem better than any analysis: in the war for elite talent, money alone no longer buys loyalty.

What does buy it, then? TechCrunch points to a structural factor: both OpenAI and Anthropic are preparing to go public, and the promise of equity that could appreciate in an imminent IPO is a magnet that an already-listed company can hardly match. Added to this are less quantifiable but equally decisive variables for a leading researcher: autonomy, access to frontier compute, and the chance to work on projects that define the state of the art. It is worth keeping the other side in view: absent official statements from Google —which TechCrunch says it contacted— this is, for now, a story told from the departures, not from the strategy of those receiving them.

For agentic AI, the move has tangible consequences. Anthropic and OpenAI are intensifying their bet on systems capable of planning, using tools and executing multi-step tasks autonomously. Bringing in profiles who have worked on a model conceived to compete in that very space could accelerate their capabilities; and Jumper's depth in structural reasoning opens doors in domains such as drug discovery or materials science, where agents are beginning to show real usefulness.

That said, it is worth resisting the temptation of the catastrophist headline. Google has historically been the world's largest incubator of AI research —from the original Transformer paper to AlphaFold itself— and has shown a notable ability to regenerate talent. The risk is not so much losing a specific scientist as the perception of an exodus feeding on itself and eroding its appeal as a destination. The underlying question is not whether Google knows how to do good research, which it does, but whether a large, consolidated corporation can compete in desirability with the combination of impact and sudden enrichment offered by a startup on its way to going public. The answer, most likely, will mark the sector's next stage.