Key Takeaways
- Four senior Google AI researchers, including Nobel laureate John Jumper, have left for Anthropic and OpenAI in recent days
- SignalFire data shows DeepMind engineers switch to Anthropic 11x more often than the reverse
- Upcoming IPOs at Anthropic and OpenAI let these companies offer equity packages Google cannot match
Google is bleeding senior AI talent. Bloomberg reports that Jonas Adler and Alexander Pritzel, both key figures behind Gemini, plan to join Anthropic. Days earlier, Nobel laureate John Jumper announced the same move, and Gemini lead Noam Shazeer left for OpenAI. Four departures in quick succession, each one a name that actually matters.
The news dragged Alphabet stock down. Investors are right to be nervous. This is not routine attrition. These researchers helped build the models Google is betting its future on.
Who left and what they built
Jonas Adler worked on AI-powered coding at Google. Alexander Pritzel focused on training AI systems. Both contributed to Gemini, Google's flagship multimodal model meant to compete with GPT-4 and Claude. Noam Shazeer, who left for OpenAI, co-authored the original Transformer paper in 2017. That paper created the architecture behind every large language model in use today.
John Jumper won the 2024 Nobel Prize in Chemistry for AlphaFold, DeepMind's protein-folding breakthrough. His departure to Anthropic is symbolically brutal. AlphaFold was supposed to be proof that Google's research lab could produce world-changing science. Now the scientist behind that proof works for a competitor.
Why they are leaving
Money. The obvious answer is almost certainly the right one. Both Anthropic and OpenAI are approaching IPOs. OpenAI's October 2024 funding round valued the company at $157 billion. Anthropic has raised $14 billion total. Early employees at either company stand to make generational wealth if those IPOs succeed.
Google cannot offer the same upside. Alphabet is already public. Its stock might grow 10% or 20% annually. A pre-IPO startup's equity can grow 10x or 100x. For a researcher who already earns $500,000 or more in base compensation, the calculus is simple. Take the risk, chase the windfall.
A SignalFire analysis quantifies the imbalance: DeepMind engineers switch to Anthropic eleven times more often than Anthropic engineers switch to DeepMind. That is not a bidirectional talent flow. It is a drain.
Google's response
Demis Hassabis, CEO of Google DeepMind, pushed back at an event in Cannes. He claimed Google has the deepest research bench of any AI lab. That may be true in raw headcount. DeepMind and Google Brain merged in 2023 to form Google DeepMind, creating one of the largest concentrations of AI researchers on the planet.
But depth does not mean stability. A lab with 1,000 researchers can lose 10 critical ones and still see its competitive position erode. Research breakthroughs often come from small teams. The question is not how many people Google employs. It is whether the people driving the next breakthrough stay.
What this means for the AI industry
The talent war has a clear winner for now. Startups with IPO equity beat incumbents with public stock. This pattern will hold until Anthropic and OpenAI actually go public. Once their shares trade freely, the equity premium disappears. They become normal companies competing on salary, culture, and research freedom.
Until then, expect more departures. Not just from Google. Meta, Microsoft, and Amazon all run large AI research groups. All face the same math. Their researchers can stay for guaranteed compensation or leave for asymmetric upside. Some will stay. The ambitious ones will leave.
Logicity's Take
For AI teams building on top of foundation models, this talent shift matters less than it looks. Gemini, Claude, and GPT-4 all work. They will keep working. The researchers who leave are building the next generation, not maintaining the current one. If you are shipping products today, pick your model based on performance and pricing, not on which lab has the most Nobel laureates. The real impact will show up in 2027 or 2028, when the models these researchers build at their new homes start shipping. By then, the competitive landscape could look very different.
Another look at how Big Tech's AI ambitions are straining internal teams
Anthropic's growing infrastructure needs as it scales up operations
Frequently Asked Questions
Why are Google AI researchers leaving for Anthropic and OpenAI?
Both Anthropic and OpenAI are approaching IPOs, allowing them to offer early-employee equity packages that could be worth far more than Google's public stock. SignalFire data shows DeepMind engineers switch to Anthropic 11 times more often than the reverse.
Which Google researchers left for competitors in 2026?
Jonas Adler and Alexander Pritzel are joining Anthropic. Nobel laureate John Jumper also moved to Anthropic. Gemini lead Noam Shazeer, co-author of the original Transformer paper, left for OpenAI.
How does this affect Google's AI competitiveness?
The departures hurt Google's ability to develop next-generation models. These researchers were central to Gemini. However, current models remain functional, so the impact on shipped products is minimal in the short term.
What did Google say about the AI talent exodus?
Google DeepMind CEO Demis Hassabis said at an event in Cannes that Google has the deepest research bench of any AI lab. He did not directly address the specific departures.
Need Help Implementing This?
Building AI products and need guidance on model selection, deployment, or team structure? Reach out to the Logicity team — we work with AI builders navigating exactly these decisions.
Source: The Decoder / Maximilian Schreiner
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
