Gemini co-lead Noam Shazeer leaves Google for OpenAI

TL;DR:

  • Noam Shazeer, a Google engineering VP and co-lead of its Gemini models, is leaving to join IPO-bound OpenAI.
  • The move comes under two years after Google reportedly paid $2.7bn (£2bn) to bring him back from his startup Character.AI.
  • Shazeer co-authored the 2017 paper that underpinned the modern AI boom.

One of the most influential figures in modern AI is switching sides. Noam Shazeer, a vice-president of engineering at Google and co-lead of its Gemini models, said he will leave to join OpenAI, the ChatGPT maker now heading towards a public listing. The departure is a notable blow given Shazeer’s role in closing the gap between Gemini and OpenAI’s own systems.

A costly hire walks out the door

The timing makes the exit sting. Google reportedly paid around $2.7bn (£2bn) less than two years ago to bring Shazeer — former head of the startup Character.AI — back into the fold along with a team of his researchers. Appointed in 2024 to help co-lead Gemini, he has been credited as a key reason the model caught up with ChatGPT. “We are grateful for Noam’s meaningful contributions,” Google said, while Shazeer called himself “excited” to join OpenAI.

His pedigree explains the bidding war over people like him: Shazeer joined Google in 2000 and co-authored the seminal 2017 “Transformer” research that catalysed today’s generative-AI surge. The episode is a vivid illustration of how the frontier labs are competing as fiercely for a small pool of senior researchers as they are for compute and customers — and how even nine-figure retention deals struggle to hold talent when a rival is preparing to float.

Looking forward

For the UK, the contest is a reminder of where the gravitational centre of frontier AI still sits. As Britain debates how to secure access to top US models and build sovereign capability, the people who actually shape those models remain concentrated in a handful of American firms. Shazeer’s move adds momentum to OpenAI ahead of its IPO and intensifies the talent squeeze — a dynamic that makes homegrown AI research capacity, and the ability to retain it, an increasingly strategic concern for the UK.

Share this article