Google Gemini co-lead Noam Shazeer leaves for OpenAI
What's the deal? Noam ShazeerDealroom has a profile for this one. Try Dealroom →, vice-president of engineering at Google and co-lead of its Gemini AI model, is leaving to join OpenAI. He will serve as OpenAI's lead for AI architecture research, focused on how to build models.
Shazeer is one of AI's most influential figures. He co-authored the landmark 2017 paper on transformer architecture — the foundation underpinning today's generative AI. He previously left Google in 2021 to found Character.AI, then returned in 2024 with a team of researchers in a deal valued at roughly $2.7B.
"I'm excited to share that I'll be joining OpenAI and look forward to working with the exceptional team there," Shazeer wrote on X. "It was a difficult decision to move on."
OpenAI chief executive officer Sam AltmanDealroom has a profile for this one. Try Dealroom → said he had wanted to work with Shazeer "since the very beginning of OpenAI," joking that it "only took 10 years" and would be "worth the wait." Google said it was "grateful for Noam's meaningful contributions" and wished him well.
Why now? OpenAI is preparing for an IPO, and landing a researcher of Shazeer's calibre adds technical credibility ahead of that transition. AnthropicDealroom has a profile for this one. Try Dealroom → is also reportedly eyeing a public listing, intensifying competition among frontier AI labs for top talent.
What could go wrong? For Google, losing a co-lead of Gemini is a significant blow during the fiercest phase of the AI race. Shazeer's institutional knowledge of Google's model architecture is hard to replace. For OpenAI, integrating a high-profile hire with strong opinions on how models should be built could create friction — or it could accelerate progress.
The signal: The AI talent war has reached the executive level. It's no longer just about poaching researchers — companies are trading their most senior technical leaders. As AI labs race toward IPOs and ever-larger models, the people who know how to build those models are the scarcest resource of all.
Read more: CNBC · Silicon Republic · Noam Shazeer on X
