Noam Shazeer leaves Google for OpenAI

Noam Shazeer, Google's VP of Engineering and a co-lead of the Gemini AI models, announced he is leaving Google to join OpenAI ^[1][2]. In his own post, Shazeer framed it as a hard exit rather than a clean break: "It was a difficult decision to move on. I'm incredibly proud of the amazing team at Google and everything we've built together." Google, for its part, acknowledged the departure and thanked him, saying it was "grateful for Noam's meaningful contributions to Google over the years" ^[2]. Notably, Shazeer's announcement did not state what his role at OpenAI will be ^[4]. Reporting circulating alongside the news has floated a specific title and a rival bidding contest, but neither has been confirmed from primary sources, and they should be treated as speculation rather than fact. What is confirmed is the headline: one of the two people steering Google's flagship model is now an OpenAI employee.

The move stings because of how recently — and how expensively — Google secured Shazeer in the first place. He originally joined Google in 2000, co-authored "Attention Is All You Need" in 2017 introducing the Transformer architecture that underpins modern generative AI, then left in 2021 to co-found Character.AI ^[1]. In August 2024 he returned as part of a reported $2.7 billion Google-Character.AI deal — a transaction the Wall Street Journal characterized as mostly a vehicle to rehire him ^[7]. Back at Google, he was made co-lead of Gemini and credited with helping close the capability gap with ChatGPT ^[1]. Leaving less than two years later turns that rehire into one of the most expensive short tenures in the industry. As one publication put it, the episode is "a reminder that the battle for AI leadership is increasingly becoming a battle for people as much as technology" ^[5].

Altman framed the hire as a long game finally paying off, calling Shazeer one of the people he most wanted to work with since OpenAI's founding and describing it as roughly a decade in the making ^[3]. For OpenAI, coverage casts the signing as a strategic win arriving just as the company has moved toward an IPO ^[5], and it has intensified speculation about the next model generation, with GPT-5.6 already being discussed as a near-term release ^[6]. The reaction among practitioners and observers is genuinely split. One camp reads it as pure acceleration — pairing Shazeer's pretraining-scaling expertise with OpenAI's existing reinforcement-learning talent to attack the areas where rivals have pulled ahead. A more skeptical camp questions the framing entirely, debating whether Shazeer was poached or quietly pushed out, and whether the destination is a company hitting its stride or one scrambling to keep pace. The truth is unknowable from the public record, but the disagreement itself is the signal: a single researcher's move is being read as a referendum on who is winning.