Key Takeaways
- Jonas Adler and Alexander Pritzel, instrumental in developing Google’s Gemini AI, are departing to join Anthropic
- Recent exits include Nobel Prize winner John Jumper (to Anthropic) and prominent researcher Noam Shazeer (to OpenAI)
- The allure of pre-IPO ownership stakes at Anthropic and OpenAI appears to be the main catalyst for these departures
- Internal resource reallocation stripped computing power from Shazeer’s initiative just before his OpenAI announcement
- Shazeer previously earned hundreds of millions when Google paid $2.7B to license Character.AI technology
The talent drain at Google’s AI division continues to accelerate. According to a Bloomberg report released Wednesday, Jonas Adler and Alexander Pritzel—both instrumental figures in the development of Google’s Gemini AI platform—are set to depart for Anthropic.
Alphabet (GOOGL) stock declined 0.30% following the announcement. These exits arrive mere days after Nobel Prize winner John Jumper made his move to Anthropic, while prominent researcher Noam Shazeer revealed his transition to OpenAI.
Adler specialized in Google’s AI-powered coding initiatives. Pritzel concentrated on the training methodologies for artificial intelligence systems. The company viewed both as essential contributors to the Gemini project.
This wave of departures has unsettled market participants and intensified concerns about Google’s capacity to retain the specialized talent necessary for maintaining its competitive edge.
A Google representative referenced statements from Google DeepMind CEO Demis Hassabis, emphasizing the organization’s continued confidence in its AI talent pipeline. Anthropic has not issued any statement on the matter.
The IPO Incentive Factor
The economic motivations driving these moves are substantial. Both Anthropic and OpenAI are widely viewed as strong IPO prospects, with potential public offerings expected between late 2026 and 2027.
For top-tier AI scientists, transitioning from a $4 trillion publicly-traded giant to a pre-IPO venture represents one of the most lucrative wealth-building opportunities available. While Google’s restricted stock units provide competitive compensation, the growth potential is largely predictable. Conversely, early equity in a rapidly expanding AI firm presents an entirely different risk-reward profile.
Shazeer’s professional trajectory demonstrates this dynamic perfectly. After leaving Google in 2021 to establish Character.AI, he later returned through a licensing arrangement valued at approximately $2.7 billion. According to the Wall Street Journal, his ownership stake in that deal generated hundreds of millions.
His latest transition to OpenAI—which recently submitted confidential IPO paperwork—potentially positions him for another significant payday if his compensation package includes equity.
Internal Resource Allocation Created Tension
Financial considerations aren’t the only factor at play. In certain instances, strategic decisions within Google seem to have amplified employee dissatisfaction.
Just before Shazeer publicized his OpenAI transition, computational resources allocated to his initiative were redirected to a Google DeepMind division based in London. Company leadership characterized this reallocation as an optimization strategy aimed at enhancing collaboration and consolidating pre-training operations—the foundational stage where AI models process massive datasets.
However, for a scientist whose work depends heavily on computational access, such a reallocation represents a significant setback. This context suggests that purely financial motivations don’t capture the complete picture.
Google spent much of the current AI race in reactive mode before building momentum in late 2025 with improved models and proprietary silicon. The consecutive departures of Jumper, Shazeer, Adler, and Pritzel now cast doubt on whether that momentum can be sustained.
Adler, Pritzel, Jumper, and Shazeer have not responded to media inquiries.


