Summary
Noam Shazeer, co-author of the foundational 2017 paper “Attention Is All You Need” that introduced the Transformer architecture, has left Google DeepMind to join OpenAI as Lead for Architecture Research. The departure, confirmed around June 18-23 by both Shazeer on X and OpenAI CEO Sam Altman, represents arguably the most significant AI talent move of 2026.
At Google, Shazeer served as Vice President of Engineering and co-lead of the Gemini model family. His return to Google in 2024 was part of a reported $2.7 billion deal involving his startup Character.AI — making his exit less than two years later a particularly bitter pill for the company. At OpenAI, Shazeer will focus on core architectural blueprints for future AI generations, bringing deep expertise in the very foundations that modern LLMs are built on.
Altman celebrated the hire by saying it “only took 10 years” to get Shazeer on board. Google issued a statement saying they were “grateful for Noam’s meaningful contributions over the years,” though the loss of a Transformer co-inventor and Gemini architect is a significant blow to DeepMind’s technical leadership.
Source
Axios · Fast Company · Business Insider
Commentary
Losing the person who literally co-invented the architecture behind your flagship product is a body blow, no matter how graciously Google frames it. Shazeer wasn’t just a symbolic hire — he was co-leading Gemini development. His departure, combined with reports of other senior researchers leaving for OpenAI and Anthropic, suggests Google’s talent retention problem in AI is accelerating despite spending billions on acqui-hires.
For OpenAI, this is a statement hire. Bringing in the person who built the Transformer to lead architecture research signals that OpenAI is thinking about what comes after the current paradigm — or at least how to push it further. Whether Shazeer can have as much impact at OpenAI as he did at Google remains to be seen, but having the Transformer architect leading your next-gen architecture research is about as good as it gets in AI recruitment.