Google Loses 2 AI Giants in 48 Hours: Nobel Winner Goes to Anthropic, Gemini Lead to OpenAI

📖 6 min read

Two of Google’s most consequential AI researchers announced departures within 48 hours this week – and they’re heading straight to Google’s two biggest rivals. Noam Shazeer, Gemini’s co-lead and a Google VP of Engineering, announced on June 17-18 that he’s joining OpenAI. Then on June 19, John Jumper – a Nobel Prize-winning chemist and VP Engineering Fellow at Google DeepMind – said he’s leaving for Anthropic after nearly nine years. Together, these departures represent the most damaging talent bleed Google has faced in the AI era.

Who Walked Out the Door

These aren’t mid-level hires. They’re architects of Google’s AI credibility.

Person	Title at Google	Key Achievement	Going To
Noam Shazeer	VP Engineering, Gemini co-lead	Co-author of “Attention Is All You Need” (the transformer paper); founded Character.AI	OpenAI
John Jumper	VP Engineering Fellow, Google DeepMind	2024 Nobel Prize in Chemistry for AlphaFold; 9 years at DeepMind	Anthropic

Shazeer is arguably one of the most influential researchers in the history of deep learning. The 2017 paper “Attention Is All You Need,” which he co-authored, invented the transformer architecture that every major AI model – GPT-4, Gemini, Claude – is built on today. He left Google in 2021 to co-found Character.AI, then returned to Google in 2024 when Google acquired the company. He lasted less than two years before announcing he’s heading to IPO-bound OpenAI.

Jumper is a different kind of loss – reputational as much as technical. He won the 2024 Nobel Prize in Chemistry alongside Demis Hassabis (Google DeepMind’s CEO) for AlphaFold, the AI system that predicted the 3D structure of virtually every known protein. That’s approximately 200 million proteins, a problem that stumped biologists for 50 years. His VP Engineering Fellow title at DeepMind puts him in a small cohort of the highest-level technical contributors at the company. He announced he’s taking a break before joining Anthropic, and has not yet revealed his specific role there.

Why This Matters Beyond the Headlines

Individual talent moves happen all the time in tech. What makes this week different is the signal-to-noise ratio.

Timing: Two high-profile departures in 48 hours is not a coincidence – it suggests a broader erosion of confidence. Senior researchers talk to each other. When one decides to leave, it lowers the psychological barrier for others.

Direction of flow: Both researchers are going to Anthropic or OpenAI – the exact companies most directly competing with Google in the foundation model market. Google is not losing talent to Meta, Microsoft, or startups in adjacent fields. It’s losing it to the companies it needs to beat.

What they represent: Shazeer embodies Google’s technical edge in language models. Jumper embodies Google DeepMind’s scientific credibility – the Nobel Prize story was central to Google’s narrative that DeepMind does real science, not just product engineering. Both narratives take a hit.

The context for Anthropic specifically: Jumper is joining a company currently embroiled in a high-stakes legal and regulatory battle in the U.S. That he’s still willing to make the move despite that environment says something about how attractive Anthropic’s research agenda looks from the inside of Google.

What Google Is Losing in Practice

Beyond prestige, these departures have real operational implications.

Shazeer’s departure from Gemini’s co-lead role is a continuity risk. Gemini has already faced public comparisons to OpenAI’s models, with benchmarks showing competitive but not dominant performance in most categories. Losing a co-lead mid-cycle is not ideal for roadmap stability.

Jumper’s work on AlphaFold touched protein biology, drug discovery, and materials science – areas where Google DeepMind has been positioning itself as the leader in AI applied to real-world science. His departure doesn’t kill that work (AlphaFold is already published and widely used), but it removes a key architect from whatever comes next.

What OpenAI and Anthropic Are Getting

OpenAI lands Shazeer just as it’s preparing for an IPO – the timing is strategic. Having a transformer co-inventor on staff is a credibility signal to institutional investors and enterprise customers alike. It also strengthens OpenAI’s research depth at a moment when the company has faced criticism for moving too fast on product and too slow on safety research.

Anthropic gets Jumper at a pivotal moment. The company raised $4 billion from Amazon in 2023 and has continued to attract major enterprise clients, but it still trails OpenAI in market share. Adding a Nobel laureate to its roster – even in an unspecified role – strengthens its claim to be the scientifically rigorous alternative in the AI race. That positioning matters for regulated industries like healthcare and pharmaceuticals, where AlphaFold’s work has the most direct relevance.

The Broader Talent War: Where Things Stand

This week’s departures are the latest chapter in a talent war that has reshaped the AI industry since 2022. Some context:

• OpenAI has now attracted talent from Google, DeepMind, Meta, and academia at an accelerating pace
• Anthropic was founded largely by ex-OpenAI researchers, and has continued to pull talent from Google
• Meta has been the outlier, using aggressive compensation packages to retain and recruit AI researchers without losing comparable figures publicly
• Google has been the primary source of outbound talent, despite paying competitively and offering world-class infrastructure

The pattern suggests compensation alone isn’t the issue. Researchers at the VP level at Google are not leaving for money. They’re leaving because of mission alignment, autonomy, or frustration with the pace and direction of work inside a large corporation.

What Google Has Going For It

This is not a eulogy for Google AI. Some honest counterpoints:

Google still has Demis Hassabis running DeepMind. It has more compute infrastructure than any rival. It has more data, a larger product surface, and more enterprise relationships. AlphaFold is still a Google product, and the institutional knowledge around it doesn’t walk out with one researcher. Gemini is still a serious product with millions of users.

Google also has a history of recovering from talent losses. The original Transformer paper team has scattered across the industry, and Transformers are everywhere – Google didn’t need to own the researchers to benefit from the architecture.

But the optics of this week are bad. And in the AI race, where perception of momentum matters nearly as much as technical reality, optics have consequences.

BetOnAI Verdict

Grade: Yellow flag for Google, green light for Anthropic and OpenAI.

This isn’t a crisis for Google – it’s a warning signal. Two world-class researchers leaving for competitors in 48 hours is a data point, not a death sentence. Google has survived bigger shake-ups. But the direction of flow matters: the best talent in AI is choosing Anthropic and OpenAI over the company with the most resources. That’s worth paying attention to.

For Anthropic, Jumper is a significant win. The company is fighting for credibility in science-adjacent markets where a Nobel Prize winner on staff is a genuine differentiator. For OpenAI, Shazeer is a pre-IPO signal that the company can attract researchers who could work anywhere. For Google, the question now is whether this is a coincidence or the beginning of a pattern – and whether Demis Hassabis has a retention plan that goes beyond compensation.

The talent war in AI is not over. But this week, Google is losing it.

---

Sources:

• Reuters – US scientist John Jumper to leave Google DeepMind for Anthropic (June 19, 2026)
• Bloomberg – Nobel Winner John Jumper to Leave Google DeepMind for Anthropic (June 19, 2026)
• Reuters – Google Gemini co-lead Noam Shazeer to join IPO-bound OpenAI (June 18, 2026)
• 9to5Google – Gemini’s co-lead is leaving Google to join OpenAI (June 17, 2026)
• Business Insider – AlphaFold pioneer who won a Nobel Prize alongside Demis Hassabis leaves Google DeepMind for Anthropic (June 2026)