Two years after Google paid billions to bring a star researcher back, he is walking out the door — exposing the weakness of buying talent.

By the numbers

Noam Shazeer is leaving Google for OpenAI — just two years after Google paid $2.7 billion to bring him and part of his team back from his start-up, Character.AI. The move, which he announced on June 18, lands like a thunderclap in an industry where a single researcher can shift the balance of power between the world’s most valuable companies.

What an “acqui-hire” is — and its limits

An acqui-hire is when a company buys another business mainly to get its people rather than its product. Google’s 2024 arrangement with Character.AI was widely read as exactly that: officially a licensing deal for the start-up’s technology, but in practice a very expensive way to bring Shazeer back into the fold and put him to work on Google’s most important AI projects. The trouble with paying for people is that people are not assets a company can lock away forever. Such deals usually come with retention packages meant to keep the talent in place for a few years — but once those periods expire, the most valuable individuals can simply leave, taking with them the very expertise the buyer paid billions to secure. Shazeer’s departure, almost exactly two years on, is a textbook illustration of that limit.

Who is Noam Shazeer

Shazeer is not just any engineer. He first joined Google in 2000 and went on to become one of the co-authors of the landmark 2017 research paper “Attention Is All You Need,” which introduced the “transformer” architecture that underpins virtually every modern AI system, from chatbots to image generators. He left in 2021 to co-found Character.AI, returned to Google in 2024, and became a co-lead of Gemini, the company’s flagship model. He is regarded as one of the strongest researchers in the field, particularly in the demanding area of pre-training — which is precisely why his decisions about where to work are studied so closely.

Why it matters

The episode is a snapshot of the AI “talent wars.” Because a handful of researchers can meaningfully change what a model is capable of, the leading labs pay extraordinary sums and poach one another relentlessly. For OpenAI, hiring Shazeer is a notable coup as it races its arch-rival Anthropic — and, according to reporting, both companies are heading toward hotly anticipated stock-market debuts, which makes every high-profile signing a marketing moment as well as a technical one. For investors, the lesson cuts deeper: when a company’s value rests heavily on a small number of brilliant people, that value is far more fragile than a balance sheet suggests, because it can quite literally walk out the door.

What to watch

Watch whether others follow Shazeer between the big labs, since star researchers often move in clusters and bring colleagues with them. Watch, too, how Google responds and whether it can hold on to the rest of its Gemini team — and how investors weigh this kind of key-person risk if and when OpenAI and Anthropic actually come to market.

Key takeaways

• Noam Shazeer is leaving Google for OpenAI, two years after a $2.7 billion deal brought him back from Character.AI.
• It exposes the limit of “acqui-hires”: once lock-ups end, star talent can walk, taking the value with them.
• The move is a coup for OpenAI in the AI talent wars as it races Anthropic toward their expected IPOs.

Source: Axios Pro Rata (Dan Primack), June 18, 2026 (The BFD), citing The Information.