The Transformer's Co-Inventor Just Left Google for OpenAI

The AI talent war just claimed its biggest name yet. Noam Shazeer, co-author of the 2017 paper that invented the Transformer and a co-lead of Google's Gemini, has left for OpenAI as its Lead for Architecture Research. Google paid 2.7 billion dollars to bring him back in 2024, and now he is gone.

The name carries real weight. Shazeer co-wrote 'Attention Is All You Need', the paper that introduced the Transformer, the architecture behind ChatGPT, Claude, Gemini and nearly every major model in use today. At OpenAI he will lead architecture research, the work of designing the next neural-network structures that the company's future models are built on. This is the person who shapes the foundation, not the features.

For OpenAI, it is a statement hire. Sam Altman welcomed him publicly, writing that Shazeer is one of the people he has most wanted to work with since the very beginning of the company, and that it only took ten years. Landing the Transformer's co-inventor as it moves toward an IPO gives OpenAI both a research edge and a recruiting signal that it is where the field's best want to be.

For Google, it stings twice over. The company paid 2.7 billion dollars in a 2024 deal to bring Shazeer back from his startup Character.AI, so losing him less than two years later is an expensive reversal. It also came in the same stretch that Google lost another AI heavyweight, with a Nobel laureate departing for Anthropic. Two of its biggest names walked out within days.

The bigger picture is an escalating arms race for people. Frontier labs are competing for a tiny pool of researchers who can actually push model architecture forward, and the price for them has reached levels that look more like sports transfers than hiring. When a single researcher can swing the trajectory of a 900-billion-dollar company, the bidding gets fierce. Talent, not just compute, is now the bottleneck.

So OpenAI just pulled the Transformer's co-inventor out of Google, and the talent war moved up a level. A landmark hire, an expensive loss, and a signal about where the momentum sits. The models run on architecture, and the people who design it are the scarcest resource in AI. Watch where the next big name lands.