Google Gemini co-lead Noam Shazeer departs for OpenAI less than two years after a $2.7bn deal brought him back

A short-lived return

Noam Shazeer, one of the most recognised researchers in artificial intelligence, is leaving Google for OpenAI. He announced the decision himself on X, writing that he was excited to join the exceptional team there and that the choice had been difficult. The move comes less than two years after Google paid roughly $2.7bn to bring him back from Character.AI, the chatbot startup he co-founded after his first exit from the company in 2021.

Shazeer was a vice president of engineering at Google and co-led the development of Gemini, the model family that powers services ranging from Google Drive and Chrome to Android smartphones. His return in August 2024, alongside co-founder Daniel De Freitas and a team of researchers, was seen at the time as a signal that Google would pay whatever it took to close the gap with OpenAI.

I'm excited to share that I'll be joining OpenAI and look forward to working with the exceptional team there.

The transformer legacy

Shazeer is one of the authors of the 2017 paper "Attention Is All You Need", which introduced the transformer architecture that underpins virtually every modern large language model, including ChatGPT, Claude and Gemini itself. His work sits at the foundation of the generative-AI boom, which makes his departure for the company Google most directly competes with more than a routine personnel change.

OpenAI gains the researcher at a moment when it is widely reported to be moving toward an initial public offering. Adding a co-inventor of the transformer is both a research hire and a statement to the market.

Friction and a pattern of exits

The departure is the latest in an aggressive talent war across Silicon Valley. Google recently lost Android security chief René Mayrhofer, and earlier moves saw OpenAI co-founders Andrej Karpathy and Durk Kingma join Anthropic. Meta, meanwhile, spent $14.3bn to secure stakes in Scale AI and its 28-year-old founder Alexandr Wang.

Shazeer’s second spell at Google was not without tension. In 2025 The Information reported that he had posted on internal channels during a transgender day of visibility, writing that he did not believe God puts people in the wrong bodies and that sterilising children was not acceptable. Jeff Dean, head of Google AI, responded with a note reminding him that leaders must support the whole employee community. Whether that friction played a role in the exit is unknown, but it adds context to a relationship that was already complicated by Shazeer’s earlier frustration in 2021, when Google declined to release his chatbot Meena to consumers.

One of the people I most wanted to work with since the earliest days of OpenAI.

What the timeline shows

Shazeer first joined Google in 2000 and spent two decades there before leaving to build Character.AI. The 2024 deal that returned him was structured as a technology-licensing agreement, but The Wall Street Journal reported that his return was widely regarded as the main reason for the $2.7bn cheque. Christopher Manning, director of Stanford’s AI lab, remarked at the time: "Noam is, without doubt, an exceptional person in this field," before asking whether anyone could be "20 times better than the rest."

Now, barely two years after that reunion, Shazeer is switching sides again. His role at OpenAI has not yet been disclosed, but the hire lands just as the contest between the two companies over enterprise AI is intensifying. At Cloud Next this year Google rebuilt its platform around agents, with Gemini central to the strategy, which makes the loss of one of its co-leads a blow that money was supposed to prevent.