Four high-profile Google AI researchers are now headed to its sharpest rivals, turning a cluster of personnel moves into a live test of whether Google can still keep the people who built its AI edge.
Jonas Adler and Alexander Pritzel are leaving Google for Anthropic, according to TechCrunch, which cited Bloomberg. Both researchers played key roles in developing Gemini, Google’s flagship model family. Their exits follow Noam Shazeer leaving Google for OpenAI and John Jumper leaving Google DeepMind for Anthropic.
The deeper signal is not that Google suddenly lacks AI talent. It doesn’t. The signal is that rival labs are now able to recruit from the core of Google’s model and science teams at the exact moment OpenAI and Anthropic are preparing to go public, giving them a cleaner equity story to sell to elite researchers.
Google AI researchers are leaving in a pattern, not a one-off
Adler and Pritzel matter because Gemini is where Google’s research credibility meets its product urgency. TechCrunch says they played key roles in the model’s development. That places their departures closer to the center of Google’s AI strategy than a routine lab reshuffle.
The timing makes it harder to dismiss. Last week, Shazeer announced he was leaving Google for OpenAI. TechCrunch described him as a legendary AI researcher who had been at Google since 2000, except for the three years he spent building Character.AI.
Google had already paid heavily to bring him back. The company effectively acquihired Character.AI for $2.7 billion, in part to return Shazeer to work on Gemini. Losing him again is not just a résumé headline. It undercuts the idea that Google can always use scale and money to reclaim key people.
Then came Jumper. He won the 2024 Nobel Prize in Chemistry with Demis Hassabis for work on AlphaFold, which can predict 3D protein structures from amino acid sequences. Reuters, as cited in the supplied material, reported Jumper’s own statement:
“After nearly nine years, I have decided to leave Google DeepMind and join Anthropic,”
That gives the latest Adler and Pritzel moves more weight. The exits cut across different kinds of AI expertise: Gemini model development, LLM history, and AI for science. One star leaving is attrition. A run of stars leaving becomes a competitive signal.
Anthropic and OpenAI are selling something Google can’t easily match
TechCrunch’s key explanation is blunt: as OpenAI and Anthropic prepare to go public, they can recruit top AI talent with the promise of equity. That is not a small perk in frontier AI. It is a weapon.
Google can pay. The source material also points to Google using special stock options to retain DeepMind researchers. But equity in a fast-rising private AI lab offers a different pitch: join now, help shape the core model, and ride the company into public markets.
XOOMAR analysis: that pitch likely works best on researchers who already have money, reputation, and optionality. For them, the question is not only compensation. It is where their work will move fastest, where it will sit closest to the model roadmap, and where the organization’s urgency matches the moment.
The supplied reporting does not state why Adler, Pritzel, Shazeer, or Jumper chose to leave. That matters. Still, the clustering supports a narrower conclusion: rivals are now credible enough to pull from Google’s most prestigious AI bench.
| Researcher | Leaving | Joining | Source-supported significance |
|---|---|---|---|
| Jonas Adler | Anthropic | Played a key role in Gemini development | |
| Alexander Pritzel | Anthropic | Played a key role in Gemini development | |
| Noam Shazeer | OpenAI | Longtime Google AI researcher, returned after Character.AI deal | |
| John Jumper | Google DeepMind | Anthropic | Co-won 2024 Nobel Prize in Chemistry for AlphaFold work |
Gemini now carries Google’s research legacy and its product pressure
Google helped create much of the modern AI foundation. The supplied material ties Shazeer to Google’s long-running AI work, Character.AI, and Gemini. It ties Jumper to DeepMind’s AlphaFold breakthrough. It ties Adler and Pritzel to Gemini itself.
That is the uncomfortable part. Google’s AI story is no longer only about invention. It is about conversion: can the company turn deep research into models, products, and customer trust quickly enough to keep both users and researchers convinced?
Gemini is the test case. It sits at the center of Google’s answer to OpenAI and Anthropic. If key Gemini contributors leave for those same rivals, the optics cut straight into the product narrative.
This does not mean Gemini is doomed. The source material does not support that claim. Google still has a huge AI organization, DeepMind, and distribution across its products. But frontier AI is unusually sensitive to small groups of exceptional people. The gap between a strong model and the best model can hinge on decisions made by a narrow set of researchers and engineers.
That product pressure extends beyond model releases. XOOMAR has separately tracked Google’s AI push across adjacent areas, including Google’s A24 AI partnership and an Android 17 Pixel 9 Pro workflow test. Those stories don’t explain these departures, but they show why the fight for Gemini talent matters across Google’s wider product surface.
Investors are treating AI talent as a balance-sheet variable
Fortune reported that the recent departures sent Google’s shares tumbling more than 5% on Monday. That reaction says investors are no longer treating elite AI researchers as interchangeable employees inside a large tech company.
The logic is simple. AI labs compete on compute, data, product distribution, safety work, and execution. But talent sits inside all of those. A researcher who shapes model architecture or scientific AI strategy can affect roadmaps, recruiting credibility, and the confidence of enterprise buyers evaluating model providers.
For Anthropic, hiring Adler, Pritzel, and Jumper strengthens the perception that it can pull from Google’s top tier. For OpenAI, hiring Shazeer adds another symbolic win against the company that helped seed the field.
For Google, the best defense is scale and depth. Attrition at a company of its size is inevitable. Not every departure signals a strategic wound. But when the names cluster around Gemini and DeepMind, the burden shifts to Google to show that its next wave of work is not walking out the door.
Cloud buyers and developers should read the exits carefully, not mechanically
Enterprise customers choosing among Gemini, Claude, OpenAI models, and other options should not overreact to personnel moves alone. Model quality, tooling, latency, safety controls, pricing, and roadmap reliability still matter more than any single hire.
Developers should watch whether the departures show up in product pace. If Gemini keeps improving and Google keeps shipping credible updates, the talent story becomes less damaging. If rival models visibly accelerate while Google looks slower, these exits will look more predictive in hindsight.
Startup investors will read the same signal differently. Researcher movement often hints at where new technical clusters may form. The supplied material already shows senior people moving from Google to Anthropic and OpenAI, while other DeepMind figures have launched startups, including David Silver with Ineffable Intelligence, according to the related Fortune source.
The mistake would be turning this into a collapse narrative. Google remains one of the deepest AI institutions in the world. The sharper question is whether its size, approval layers, and product risks make it less attractive to researchers who want direct impact on frontier systems.
Google’s next AI recruiting fight will be won by proof, not perks
The near-term evidence to watch is concrete: whether Google can retain senior Gemini and DeepMind figures, whether Anthropic and OpenAI keep hiring from the same pool, and whether Gemini’s public progress convinces developers that the core team remains strong.
A stronger Google response would likely involve more than richer pay. It would need visible autonomy for elite AI teams, faster paths from research to product, and a clear reason for top researchers to believe their best work can still happen inside Google.
A weaker signal would be more clustered exits from Gemini, AlphaFold-adjacent science teams, or senior model leadership, especially if those researchers land at Anthropic or OpenAI before public offerings.
Google will not lose the AI race because Adler, Pritzel, Shazeer, and Jumper left. But if the pattern keeps telling researchers that the most urgent work is happening elsewhere, Google’s problem becomes cultural as much as competitive.
The Stakes
- Google is losing researchers from the core teams behind Gemini and AlphaFold.
- Anthropic and OpenAI are proving they can recruit elite talent directly from Google’s AI bench.
- Potential IPO equity upside may be making rival labs more attractive to top researchers.
Originally published on XOOMAR. For more news and analysis, visit XOOMAR.
Top comments (0)