Google DeepMind spent years as the undisputed magnet for Europe’s top machine-learning minds. That grip is now slipping fast. New data shared exclusively with Fortune shows the lab’s share of research and advanced-engineering hires across Europe, the Middle East and Africa has collapsed from 49% in 2022-23 to just 18.6% in 2025-26, the sharpest regional market-share drop recorded for any major AI lab. The shift in Google DeepMind AI talent flows marks one of the clearest signs yet that the company once seen as the industry’s research powerhouse is losing the war for the people who actually build frontier models.
Key takeaways
- DeepMind’s share of EMEA research and advanced-engineering hires fell from 49% in 2022-23 to 18.6% in 2025-26, according to data from Zeki Data.
- DeepMind’s arrival-to-departure ratio dropped from about 12-to-1 in mid-2023 to roughly 2-to-1 by the third quarter of 2026, versus 22-to-1 at Anthropic in 2025.
- Jeff Dean, Sanjay Ghemawat, Oriol Vinyals and Quoc Le departed this month, while co-founder Demis Hassabis stepped back from day-to-day control of the lab.
- Of DeepMind staff who left in the past 12 months, 25% moved to Anthropic, 21% to Meta and 14% to OpenAI.
- DeepMind is losing ground in large language models, multimodal systems and computer vision, but still leads in robotics, embodied AI and scientific machine learning.
DeepMind’s Shrinking Share of Europe’s AI Talent Pool
The numbers tell a story of a lab that once had the field to itself and now shares it with much hungrier rivals. Zeki Data, a UK-based data intelligence company, tracked research and advanced-engineering roles across companies using public records. Among the findings: DeepMind’s EMEA hiring share fell significantly in just three years, while OpenAI and Anthropic pulled ahead in the race for the same pool of specialists.
“They had the crown in Europe forever, and then it started to erode from a very high base,” Tom Hurd, founder of Zeki Data, told Fortune. “The likes of Microsoft AI Superintelligence and Meta Superintelligence are eating into their market share, and then there’s OpenAI and Anthropic on the side.”
A Widening Gap in Retention
Retention data paints an even starker picture. DeepMind’s arrival-to-departure ratio — a measure of how many people it hires for every one who leaves — fell from roughly 12-to-1 in the second quarter of 2023 to about 2-to-1 in the third quarter of 2026. That means the lab is now bringing in only two research and engineering hires for every departure, compared with twelve just three years earlier.
By contrast, Anthropic posted a 22-to-1 ratio in 2025, while Meta stood at 3-to-1 and OpenAI at 5.7-to-1, according to Zeki’s report. Hiring growth tells a similar story: DeepMind’s research and engineering headcount has grown at roughly 27% annually since 2022, compared with 97% at OpenAI and 152% at Anthropic — a gap that reflects how much faster the newer labs are scaling relative to their smaller starting base.
Leadership Upheaval Adds to the Pressure
The talent drain isn’t limited to rank-and-file researchers — it has reached DeepMind’s top leadership. This month alone, the lab lost Jeff Dean, its long-serving chief scientist and a 27-year veteran of the company, along with senior fellow Sanjay Ghemawat and researchers Oriol Vinyals and Quoc Le, who left together to launch a startup called Discovery Loop.
On the same afternoon, co-founder and chief executive Demis Hassabis announced he would step back from day-to-day control of the lab, becoming chairman while also taking on the role of Alphabet’s chief scientist. Operational control now falls to chief technology officer Koray Kavukcuoglu. The timing of these exits, arriving alongside the market-share and retention figures, underscores why the broader story about Google DeepMind AI talent losses is drawing attention across the industry — this isn’t just junior researchers chasing bigger paychecks, it’s the people who defined the lab’s identity stepping away.
Why Researchers Are Leaving — And Where They’re Going
Interviews with current and former staff point to a mix of factors: aggressive poaching by rivals offering cash-heavy packages, frustration inside Google over its position in the AI race, sinking morale, and the pull of pre-IPO equity at competitors. Three current and two former DeepMind employees described the exodus to Fortune in similar terms — a lab increasingly organized around commercializing Gemini rather than pursuing the open-ended science that first attracted many of its researchers.
Of the people who left DeepMind in the past 12 months, 25% went to Anthropic, 21% to Meta, and 14% to OpenAI, Zeki’s data shows. Anthropic has become the single leading destination for departing DeepMind talent.
The list of high-profile departures keeps growing. David Silver, the reinforcement-learning pioneer behind AlphaGo, AlphaZero and AlphaStar, left after nearly 13 years to launch Ineffable Intelligence, a London startup now valued at $5.1 billion following a $1.1 billion seed round. Other veterans who departed include Wojciech Czarnecki, now CTO at Fundamental, and Lasse Espeholt.
Publication Rules Add Friction for Open Research
DeepMind’s retreat from open publication appears to have compounded its talent problem. The lab tightened its internal review process and introduced a six-month embargo for some strategically sensitive generative-AI papers, a change first reported in April 2025, as it sought to keep competitors from benefiting from its research. DeepMind said at the time it remained committed to publishing and was simply updating its policies.
For a lab built on public milestones like AlphaGo and AlphaFold, the shift created friction with researchers who joined expecting relatively unconstrained work. As one person familiar with the culture put it, “Most old timers who joined DeepMind before the ChatGPT moment, joined to be part of an AI research lab, and suddenly they were asked to build products for Google.” Hurd said the tighter publication rules coincided with — and may have contributed to — the deterioration in talent flows.
Where DeepMind Still Leads
The losses aren’t spread evenly across research areas, and that unevenness matters for understanding where the competitive threat is sharpest. Zeki’s data shows a net loss for DeepMind in large language models and multimodal systems: 19.2% of departing staff specialized in those fields, compared with just 15.6% of new hires. The lab has also lost ground in computer vision.
DeepMind is gaining ground, however, in robotics, embodied AI and machine learning for science — areas where it increasingly competes for talent with Nvidia as much as with rival AI labs. The company also faces new competition outside its traditional strongholds: Mistral AI and Anthropic have been the main beneficiaries of its EMEA decline, while in Asia-Pacific, where DeepMind opened a research facility in Singapore during November, local competitors such as ByteDance, Sakana AI and Sarvam AI are expanding their presence in markets the established labs once underinvested in.
Google DeepMind still has Alphabet’s compute, cash and institutional reach behind it — resources few competitors can match. But in a market where top researchers can pick almost any lab and work on almost any problem, that reach alone no longer guarantees loyalty the way it once did.
FAQ
Why is Google DeepMind losing AI talent to other labs?
DeepMind is losing talent due to aggressive poaching by rivals offering high pay and equity, a shift from open research toward commercialization, morale issues, and tighter research publication policies.
Which companies are top destinations for departing DeepMind employees?
Departing DeepMind talent primarily moves to Anthropic (25%), Meta (21%), and OpenAI (14%), according to Zeki Data.
What areas of AI research is DeepMind losing ground in?
DeepMind is losing ground especially in large language models, multimodal systems, and computer vision.
What strengths does DeepMind retain despite talent loss?
DeepMind remains strong in robotics, embodied AI, and scientific machine learning, areas where it is still expanding its research capacity.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
Be the first to comment