Google AI leader Noam Shazeer leaves for OpenAI, intensifying talent war

Noam Shazeer’s jump from Google to OpenAI is more than a high-profile hire. It is a reminder that the AI race is now being fought on two fronts: model performance and the rare engineers who know how to build those systems at scale.

Shazeer was vice president of engineering at Google and co-led development of Gemini, making him one of the most consequential figures inside the company’s AI effort. His departure lands at a sensitive moment for Google, which has been trying to narrow the gap with OpenAI while pushing Gemini 3.5 and new agent tools such as Antigravity. Losing a leader tied so closely to Gemini is a symbolic blow, especially because Google paid about $2.7 billion less than two years ago to bring Shazeer back, along with a team from Character.AI.

The talent stakes are especially high because Shazeer is not just an executive manager. He joined Google in 2000 and was a co-author of the 2017 paper Attention Is All You Need, the work that introduced the Transformer architecture and helped ignite the modern AI boom. That paper’s author list also included Ashish Vaswani, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser and Illia Polosukhin. In an industry where a handful of technical minds helped define the current era, his move represents a transfer of institutional knowledge as much as a change of employer.

OpenAI is pressing its own advantages while it prepares for a larger public market debut. The company confidentially filed a draft S-1 with the SEC on June 8, 2026, and Reuters reported on June 15 that OpenAI spent $34 billion last year. It also announced the acquisition of Ona on June 11 to strengthen Codex with secure cloud execution for long-running agents. Those moves show a company expanding product depth and infrastructure at the same time it is trying to keep its lead in model capability and developer mindshare.

For Google, the departure underscores how expensive and unstable elite AI talent has become. For OpenAI, it adds a veteran builder with direct experience shaping one of the industry’s most important model families. The broader message is plain: in 2026, the competition between leading AI labs is no longer just about compute and code, but about who can retain the few people capable of turning both into durable advantage.