MUMBAI: DeepMind’s talent pool is getting a little less deep. New data cited by Fortune suggests Google’s AI research lab is losing ground to rivals in the race for elite researchers, with OpenAI, Anthropic and Meta increasingly drawing talent that once might have headed its way.
An analysis by UK-based data intelligence firm Zeki Data found that DeepMind’s share of research and advanced-engineering hires across Europe, the Middle East and Africa fell sharply from 49 per cent in 2022-23 to 18.6 per cent in 2025-26. Zeki described this as the steepest decline it has recorded for a major AI laboratory in any region.
DeepMind is still hiring more researchers and advanced engineers than it loses, but the gap has narrowed dramatically. Its arrivals-to-departures ratio fell from around 12-to-1 in the second quarter of 2023 to about 2-to-1 in the third quarter of 2026, according to the data.
The talent tug-of-war is particularly visible in where former DeepMind employees are going. Among researchers and advanced engineers who left the company over the past year, 25 per cent moved to Anthropic, 21 per cent joined Meta and 14 per cent went to OpenAI.
The rival labs, meanwhile, have been pulling in talent at a faster clip. Zeki’s data puts the arrivals-to-departures ratio at 3-to-1 for Meta in 2025, 5.7-to-1 for OpenAI and 22-to-1 for Anthropic.
The movement includes some high-profile names. Jeff Dean, DeepMind’s long-time chief scientist, left after 27 years and joined Sanjay Ghemawat, Oriol Vinyals and Quoc Le in launching startup Discovery Loop. Researchers David Silver, Wojciech Czarnecki and Lasse Espeholt have also departed, while Misha Laskin and Ioannis Antonoglou left in 2024 to establish Reflection AI.
The exits have reached the heart of DeepMind’s generative AI work. Noam Shazeer, a Gemini co-lead, left for OpenAI in June, while John Jumper, who shared the 2024 Nobel Prize in Chemistry with Demis Hassabis for work on AlphaFold, moved to Anthropic. Other AlphaFold researchers, including Jonas Adler and Alexander Pritzel, subsequently followed Jumper.
Zeki found that 19.2 per cent of people leaving DeepMind over the past 12 months specialised in large language models and multimodal systems, compared with 15.6 per cent of incoming talent. The analysis also pointed to losses in computer vision.
AlphaFold, one of DeepMind’s most celebrated research achievements, has experienced notable turnover too. Of the 29 named authors of the AlphaFold2 paper, 13 have since left DeepMind, according to Zeki.
The talent shifts do not mean DeepMind is retreating across the board. The lab has continued to attract researchers working in robotics, embodied AI and machine learning for science, while its wider research portfolio spans Gemini, enzyme design, genomics and nuclear fusion.
But the hiring pressure comes as DeepMind’s research culture is also changing. The company has introduced a more stringent internal review process and a six-month embargo for some strategically sensitive generative-AI papers, according to a Financial Times report cited in the supplied material. DeepMind has said it remains committed to publishing research while updating policies to protect its teams’ ability to contribute to the wider research community.
Current and former employees cited by Fortune pointed to several factors behind the departures, including aggressive recruitment by rivals, concerns about Google’s position in the AI race, lower morale and the attraction of pre-IPO equity at companies such as OpenAI and Anthropic.
Leadership has shifted too. Demis Hassabis has stepped away from day-to-day management to become chairman of DeepMind and Alphabet’s chief scientist, while CTO Koray Kavukcuoglu has taken operational control.
For a laboratory that helped define modern AI research, the challenge now is not simply producing the next breakthrough. It is making sure the people capable of producing it still want to stay.