Google DeepMind has lost four prominent AI researchers to its direct competitors in ten days, and the specific accumulation of exits is what distinguishes this from normal talent movement in a competitive sector. Jonas Adler, who led work on Google’s AI coding efforts and contributed to AlphaFold research, and Alexander Pritzel, whose focus was pretraining – the foundational stage where models learn from large data volumes – are both leaving for Anthropic, adding to the already-reported departures of Noam Shazeer to OpenAI and John Jumper to Anthropic. Both Adler and Pritzel also contributed to AlphaFold alongside Jumper, creating a striking situation where multiple members of the team behind one of AI research’s most celebrated recent breakthroughs are now moving to the same competitor. Pre-IPO equity at Anthropic is the mechanism drawing them, and it is the pull that NewsTrackerToday picks up as the structural force behind a talent wave Google cannot neutralize with compensation alone.
The specific capabilities Adler and Pritzel represented at Google matters for understanding the competitive stakes. Adler worked on AI coding at a moment when Claude Code has become one of Anthropic’s most commercially significant products and the platform that Boris Cherny, its creator, described at Meta’s @Scale conference as representing the next frontier of agentic AI. Pritzel worked on pretraining, which is the foundational work that determines what a model learns before any fine-tuning or instruction-following is applied. Both domains – coding applications and pretraining fundamentals – are ones where Anthropic is actively building competitive advantage against Google’s Gemini. Andrej Karpathy, a co-founder of OpenAI who later joined Tesla, also reportedly joined Anthropic in May 2026, adding further pre-IPO research credibility.
Sophie Leclerc, who covers the technology sector, reads the retention problem with structural clarity: “Google’s challenge is specific: it has the resources to compensate researchers well, but it cannot offer the combination of research autonomy, concentrated mission focus, and pre-IPO equity that Anthropic and OpenAI offer at this moment. Pre-IPO equity at a company approaching $1 trillion in valuation that is expected to go public before year-end is, in simple dollar terms, a potentially transformative financial event for someone joining now. The question of whether Google can restructure its AI research environment to offer comparable autonomy without the equity event is what I’d call genuinely open. Spinning off DeepMind into a more independent entity has been discussed internally, but that kind of structural change takes years, not months.” The IPO equity mechanism is what NewsTrackerToday maps as the pull that Alphabet cannot currently replicate for researchers already at a public company.
Alphabet shares closed down approximately 5% on Monday following the Shazeer and Jumper announcements, and the stock had not fully recovered by Tuesday when the Adler and Pritzel reports emerged. The 5% decline represents roughly $90 billion in market capitalization erased in a day on the basis of four individual departures, which signals that investors are not reading these exits as noise. They are reading them as a signal about Google’s ability to maintain competitive position in model development specifically. Microsoft CEO Satya Nadella added pressure in a Sunday interview calling for less dependence on “AI Giants” and suggesting AI models were being commoditized – a comment that, taken alongside the researcher departures, created a bad-news cluster that hit Alphabet harder than any individual item would have.
John Jumper is the most symbolically significant departure. Alongside DeepMind CEO Demis Hassabis, he won the 2024 Nobel Prize in Chemistry for AlphaFold, the protein structure prediction system that has had more practical impact on biological and medical research than any AI system in recent memory. That AlphaFold’s co-creator has decided to move to Anthropic at this specific moment is what NewsTrackerToday reads against the Alphabet share price decline: the market is treating the Nobel laureate’s departure as evidence that Google DeepMind’s gravitational pull on exceptional research talent is weaker than its historical reputation would suggest.
Can Google maintain its technical edge in large language model development if its best researchers keep leaving? The honest answer is that two waves of departures – the 2021 wave that produced Character.AI, OpenAI’s current architecture team, and several frontier startups, and the current 2026 wave heading to Anthropic and OpenAI specifically – have not prevented Google from building competitive Gemini models. The products still work. But each departure narrows the specific talent advantage that justifies Alphabet’s $141 billion in debt and equity raised for AI since October. The question that News Tracker Today names as the one Wednesday’s exits force is whether Google’s institutional resources can substitute for the individual researchers it loses, or whether there is a compounding effect where each departure makes the next one easier to justify.
