Skip to content
News Leadership

Transformer Co-Author Noam Shazeer Leaves Google to Join OpenAI as Architecture Research Lead

· by Pondero Newsdesk

The short version

Noam Shazeer, a co-author of the 2017 Transformer paper and former Gemini co-lead at Google, announced on June 18 that he is joining OpenAI to head architecture research.

Transformer Co-Author Noam Shazeer Leaves Google to Join OpenAI as Architecture Research Lead

Noam Shazeer announced on June 18 that he is leaving Google to join OpenAI, where he will lead architecture research. The move strips Google of one of its most prominent Gemini figures less than two years after Google paid approximately $2.7 billion to bring him back.

What

Shazeer is a co-author of "Attention Is All You Need," the 2017 paper that introduced the Transformer architecture now at the core of nearly every large language model in production. He spent decades at Google before leaving in 2021 to co-found Character.AI, the AI role-playing startup. In August 2024, per TechCrunch, Google rehired him through a deal that gave Google access to Character.AI technology for around $2.7 billion, and he returned as a co-lead of the Gemini model program.

He posted his departure on X on June 18. At OpenAI, his title will be Lead for Architecture Research, focused on next-generation model design.

The same week, former Trump administration AI policy official Dean Ball announced he will join OpenAI on July 6 to lead a new "Strategic Futures" team, per his own post on X. Ball will report to Chief Strategy Officer Jason Kwon. His team's stated scope covers catastrophic risk, recursive self-improvement, labor market effects, and the company's relationship with the U.S. government, per TechCrunch's reporting on his blog post.

Why it matters

Shazeer's arrival at OpenAI is a concrete signal of the lab-to-lab talent movement accelerating ahead of OpenAI's planned IPO. For Google, the departure of a co-lead of its flagship model program is an unusually public setback. Google had structured the Character.AI acquisition specifically to return Shazeer to Gemini development. That the arrangement lasted under two years underscores how competitive compensation and research scope have become between the top labs.

At OpenAI, Shazeer's focus on architecture is distinct from fine-tuning or product work. Architecture choices set the ceiling for what a generation of models can do. His involvement at the research level signals OpenAI is investing in the underlying design of its future model series, not only in scaling existing approaches.

Ball's hire adds a complementary dimension. As OpenAI approaches a public offering, having someone with direct White House policy experience running an internal governance team reflects an awareness that IPO investors will scrutinize regulatory exposure. The dual announcements in one week suggest a coordinated pre-IPO positioning effort rather than coincidence.

Context

Shazeer's move is the latest in a round of high-profile departures from Google's AI division in 2025 and 2026, which have also seen researchers move to Anthropic, Meta, and independent labs. Sam Altman publicly welcomed both hires, per TechCrunch. Google has not announced a Gemini co-lead replacement as of June 19. The dual announcements also landed alongside news that Anthropic faced a U.S. export control order affecting two Claude model variants, per TechCrunch, sharpening the contrast between the two labs' positions heading into the second half of 2026.

What to watch next

Google replacing or not replacing Shazeer in the Gemini co-lead role will signal how the company manages frontier-research continuity. Ball's Strategic Futures team starts July 6; its first public output will define whether the mandate translates into meaningful policy positioning for OpenAI or stays largely internal.

Sources