Noam Shazeer, one of the key leaders behind Google’s Gemini AI models, is leaving Google’s Gemini team for OpenAI. The move is awkward timing for Google: it comes shortly after fresh Gemini launches and once again spotlights how aggressively AI labs are fighting over the people who build the products, not just the products themselves.
Shazeer said he was ”excited” to join OpenAI and praised Google’s team, while also acknowledging that leaving was not an easy decision. That’s standard executive farewell language, but the part that matters is simpler: OpenAI just landed a researcher with deep experience in generative AI, and Google loses one of the names most closely tied to its current model strategy.
A fast return, then another departure
The twist here is how quickly the exit happened. Shazeer left Google in 2021 with Daniel De Freitas to found Character.AI, then both returned to Google in August 2024 as part of DeepMind. Less than two years later, Shazeer is out again. In AI, loyalty tends to last exactly as long as the next labs check clears.
That kind of movement is becoming a feature of the sector, not an exception. Meta, OpenAI, and Google have all spent the past two years trying to lock in top technical talent as model development gets more expensive, more competitive and more dependent on a handful of senior engineers who can steer research and product together.
Google’s Gemini push loses a familiar face
The departure lands just weeks after Google introduced new AI products including Gemini 3.5 Flash and Gemini Spark. That makes the timing especially messy: launches are supposed to signal momentum, not raise questions about who is still in the room when the next version gets built.
- Former role: vice president of engineering at Google and co-leader of Gemini model development
- Next stop: OpenAI
- Recent Google products: Gemini 3.5 Flash and Gemini Spark
For Google, the bigger issue is not one person leaving; it is the signal that even after a high-profile return, the company still cannot fully anchor its AI talent. Expect more of this. The hottest names in generative AI are becoming mobile assets, and every major lab now has to defend itself against the same problem: the competition is no longer just for users, but for the people who know how to make the models work.