Google Loses Two AI Titans in 24 Hours: Shazeer to OpenAI, Jumper to Anthropic

In the history of the AI industry, no single institution has been more central to the foundational research that made modern AI possible than Google. In a 24-hour window in mid-June 2026, that institution lost two of its most consequential scientists to its most aggressive rivals.

On June 18, Noam Shazeer — co-inventor of the Transformer architecture, VP of Engineering at Google, and co-lead of the Gemini model family — announced he is leaving Google to join OpenAI. The following day, John Jumper — lead architect of AlphaFold, winner of the 2024 Nobel Prize in Chemistry, and a researcher who spent nearly nine years at DeepMind — announced he is leaving for Anthropic.

The two departures, announced within less than 24 hours of each other, constitute the most significant simultaneous loss of foundational AI talent from any single organization in the industry’s history.

Noam Shazeer: The Architect of the Architecture

It is difficult to overstate Noam Shazeer’s importance to modern AI. In 2017, as co-author of “Attention Is All You Need,” he was one of eight researchers who introduced the Transformer architecture that now underlies virtually every large language model in existence — GPT, Claude, LLaMA, Gemini, Grok, and hundreds of others.

After a brief departure to co-found Character.AI, Shazeer returned to Google in 2024 when the company struck a licensing deal worth approximately $2.7 billion — widely interpreted as paying effectively to bring him back into the fold. He became VP of Engineering and was co-leading Gemini development.

Less than two years later, he is leaving again.

Shazeer announced his move to OpenAI on X on June 18, writing: “I’m excited to share that I’ll be joining OpenAI and look forward to working with the exceptional team there. It was a difficult decision to move on. I’m incredibly proud of the amazing team at Google and everything we’ve built together.”

OpenAI CEO Sam Altman’s response was uncharacteristically effusive: “Only 10 years in the making. One of the people I have most wanted to work with since the very beginning of OpenAI.”

At OpenAI, Shazeer’s mandate is reported to be next-generation architectures — a signal that OpenAI is looking beyond incremental improvements to its existing GPT stack toward fundamentally new model designs. With the architect of the Transformer now working on what comes after the Transformer, the implications for the field are significant.

A striking historical footnote: with Shazeer’s departure, all eight co-authors of “Attention Is All You Need” have now left Google.

John Jumper: The Nobel Laureate Who Solved Protein Folding

John Jumper’s contribution to science is of a different order than even Shazeer’s. As the lead architect of AlphaFold 2, Jumper solved what had been one of biology’s most intractable grand challenges: predicting the three-dimensional structure of proteins from their amino acid sequences with near-experimental accuracy.

The implications for drug discovery, materials science, and molecular biology were immediate and profound. AlphaFold’s protein structure database now covers the vast majority of known proteins and has accelerated research timelines across dozens of disease areas. In recognition, Jumper — along with AlphaFold’s scientific advisor Demis Hassabis and biochemist David Baker — was awarded the 2024 Nobel Prize in Chemistry.

After nearly nine years at DeepMind, where AlphaFold was built, Jumper announced on June 19 that he is leaving for Anthropic. His X post was characteristically understated: he plans to take time to recharge before beginning in his new role, and no title or team assignment has been disclosed.

But the strategic logic of Anthropic’s recruitment is clear. The company has been executing an aggressive AI-for-science buildout throughout 2026: opening wet laboratories, publishing research on AI agents designed specifically for biological experimental workflows, announcing flagship research partnerships with the Allen Institute for Brain Science and the Howard Hughes Medical Institute, and shipping Claude Science — the AI tool for automated multi-step scientific workflows unveiled last month.

Jumper’s expertise sits at exactly the intersection Anthropic is targeting: deep biological knowledge, computational prowess at the frontier of AI capability, and practical experience translating AI breakthroughs into tools that accelerate scientific discovery at scale.

The Competitive Calculus

These two departures, considered together, illuminate the current competitive dynamics of the AI talent market in ways that quarterly revenue figures cannot.

Google has, for over a decade, attracted and retained AI research talent partly through scientific prestige — the ability to work on breakthrough research at scale, with unmatched compute resources and an exceptional peer group. But the industry has matured. OpenAI, Anthropic, and others now offer comparable or superior research environments, substantially larger compensation packages, and — crucially — the sense that the most impactful work is being done inside the startup ecosystem rather than inside a large corporation with competing priorities.

For OpenAI, hiring Shazeer is also strategically timed. The company is preparing for what is expected to be a landmark IPO, and Shazeer’s name — more than almost any other in AI — carries credibility with the technical community. Sam Altman’s “10 years in the making” comment suggests this was a long-pursued target, not an opportunistic hire.

For Anthropic, the Jumper hire is a statement about the company’s long-term scientific ambitions. Claude Fable 5’s general AI capabilities are increasingly validated at the frontier, and Anthropic’s next act appears to be staking out a leadership position specifically in scientific AI — an area where AlphaFold’s legacy gives Jumper unmatched institutional credibility.

What Remains at Google

Google DeepMind is not suddenly diminished. Demis Hassabis leads the organization and remains among the most visionary figures in AI. Veo, Imagen, Gemini’s Deep Think mode, and the company’s robotics and healthcare AI programs represent genuinely frontier work. Google’s compute advantages, given its TPU investments and data center scale, remain unmatched by any competitor other than Microsoft.

But the departure of all eight original Transformer authors from Google is a symbolic inflection point that resonates beyond the specific individuals involved. It signals that the center of gravity of AI research — where the most ambitious scientists choose to direct their careers — has shifted, perhaps permanently, away from the organization that did more than any other to create the field.

The question for Google is not whether it can continue to do great AI research. It clearly can. The question is whether it can regain the ability to attract and retain the category of talent that defines entire epochs — the Shazeers and Jumpers — or whether it must be content competing on execution rather than foundational discovery. Based on the last 24 hours of June 2026, that question does not yet have a comfortable answer.