Transformer Co-Author Noam Shazeer Leaves Google, Joins OpenAI

Noam Shazeer, one of the core authors of the Transformer architecture, announced in June 2025 that he is leaving Google to join OpenAI as Head of Architecture Research. OpenAI CEO Sam Altman said Shazeer has been one of the people he most wanted to work with since the company's founding.

Background and Contributions

Shazeer is a co-author of the seminal 2017 paper "Attention Is All You Need," which introduced the Transformer architecture that underpins modern large language models. He also contributed to key technologies such as Mixture of Experts (MoE), Mesh-TensorFlow, and T5, profoundly influencing the development of modern AI.

Two Departures from Google

Shazeer joined Google in 2000. In 2021, he left after Google refused to release a chatbot he had developed, co-founding Character.AI with Daniel De Freitas. In 2024, Google brought him back through a technology licensing agreement worth approximately $2.7 billion and appointed him co-technical lead for Gemini. However, less than two years later, he chose to leave again.

Reactions

OpenAI's Head of Research, Mark Chen, stated that Shazeer's work on Transformers, MoE, and efficient decoding has shaped modern AI, emphasizing his belief in AGI. Shazeer himself said leaving Google was a difficult decision but expressed gratitude to the Google team.

Impact

Shazeer's addition brings together two original Transformer authors at OpenAI (the other being Lukasz Kaiser), marking a major victory in the AI talent war. For Google, the Gemini project loses a key figure, potentially affecting its model development progress.