Google DeepMind's Share of Elite AI Talent Market Fell From 49% to 18.6%

Google DeepMind is losing its lead in the race for top artificial intelligence researchers. Analytics firm Zeki Data has published data showing the lab's share of elite AI talent hires across Europe, the Middle East and Africa fell from 49 percent in 2022-23 to 18.6 percent in 2025-26.

What the numbers show

Zeki Data is a UK-based firm that tracks talent flows across the tech industry. Its analysis covered hiring and departures among the most sought-after AI researchers and engineers. The result is unambiguous: where OpenAI and Anthropic are steadily gaining market share in talent, DeepMind, and to a lesser extent Meta, are losing ground.

The most telling figure is the drop in the hire-to-departure ratio. As recently as mid-2023, DeepMind hired twelve new employees for every one who left. Three years later, that ratio stands at just two to one, meaning the company has nearly stopped replacing its losses with new hires. Anthropic, over the same period, reached a ratio of 22 to 1, meaning it is growing far faster than it is losing staff.

An August wave of departures

August 2026 alone brought a string of high-profile exits from DeepMind. The lab lost Jeff Dean, Google's longtime chief scientist and a 27-year veteran of the company, along with senior fellow Sanjay Ghemawat. Researchers Oriol Vinyals and Quoc Le also left, going on to launch their own startup, Discovery Loop.

These are not the first names on the list, and likely not the last. David Silver, a pioneer of reinforcement learning and co-creator of AlphaGo, founded his own company, Ineffable Intelligence, already valued at $5.1 billion. John Jumper, a Nobel laureate for AlphaFold, joined Anthropic in June 2026, along with researchers Jonas Adler and Alexander Pritzel. Noam Shazeer, co-lead on Gemini and co-creator of the transformer architecture, moved to OpenAI around the same time.

Why they are leaving

Fortune, which first reported the Zeki Data figures, spoke with three current and two former DeepMind employees. They pointed to several overlapping causes: aggressive, cash-heavy offers from rivals like Meta and Microsoft, mounting frustration over Google's position in the AI race, sagging team morale, and the appeal of pre-IPO equity at competitors including OpenAI and Anthropic.

They've had the crown in Europe forever, and then it started crumbling from a very high level - Tom Hurd, founder of Zeki Data

A change at the top

The departures coincide with a leadership reorganization. Demis Hassabis, DeepMind's co-founder and longtime CEO, has stepped back from day-to-day management of the lab to become chairman and chief scientist of Alphabet. Operational control of DeepMind has passed to chief technology officer Koray Kavukcuoglu.

The exodus has also hit specific research teams hard. Of the 29 authors of the landmark AlphaFold2 paper, the model that revolutionized protein structure prediction and earned DeepMind a Nobel Prize, 13 have already left the company. In large language models and multimodal systems, DeepMind has logged more departures than new hires over the past year, and the same pattern holds in computer vision.

What it means for the industry

The loss of elite researchers matters beyond Google itself. Teams that built breakthrough technologies such as AlphaFold, AlphaGo and the transformer architecture underpinning today's language models are scattering across rival labs and new startups, accelerating the pace of innovation industry-wide while eroding Google's long-standing edge in fundamental research.

For investors and enterprise customers, the Zeki Data figures signal that Google's technological edge in AI is not guaranteed, despite the scale of computing power and data at Alphabet's disposal. Holding onto that position will depend on whether the new leadership under Kavukcuoglu can stem the outflow of talent before it translates into visible delays in future Gemini models.