Analysis
China now hosts 40.6% of the world's top AI researchers, up from 27.1% in 2022, according to an updated Global AI Talent Tracker from the Carnegie Endowment's Carnegie China program, corroborated by The Information and findings from this year's Stanford AI Index.
How The Study Measures 'Top Talent'
Carnegie's methodology tracks authors publishing at NeurIPS, ICML and ICLR -- the three most prestigious AI/machine-learning research venues, where acceptance is widely treated as a marker of elite research capability -- and identifies each researcher's undergraduate institution as a proxy for country of origin, alongside their current institutional affiliation. That approach measures where top researchers currently work, not simply where they were trained, which is what makes the shift from 27.1% to 40.6% a genuine relocation-and-retention story rather than just a reflection of China producing more STEM graduates overall.
The Numbers Behind The Headline
Chinese AI talent migration to the US has fallen 89% since 2017 -- far fewer researchers trained in China are choosing to build careers at US institutions and companies than a decade ago. At the same time, 87% of Chinese-origin researchers already established in the US have stayed despite rising geopolitical tension between the two countries, meaning this is primarily a story about where the next generation of researchers chooses to go, not a mass departure of those already here. Of the 20 US-based researchers in the global Top 100 list, half are of Chinese descent, including Carnegie Mellon's Zhu Junyan and Stanford's Fei-Fei Li -- a reminder that the talent competition isn't a binary US-versus-China split so much as a question of where researchers of any background choose to work.
The Gap China Hasn't Closed
The report also identifies a structural weakness inside China's own system: fewer than 15% of Chinese university AI researchers have industry ties, compared with 37% of researchers at a peer institution like Carnegie Mellon. That 'invisible wall' between academia and enterprise means China's research talent advantage has not yet fully converted into the same tight academic-industry feedback loop that has powered US AI commercialization -- a gap that could narrow or persist depending on how China's AI industry chooses to engage with its universities going forward.
What This Means For US Policy And Investors
The finding complicates a simple containment narrative in US AI policy: talent migration restrictions and export controls were partly premised on the idea that keeping top researchers in the US would preserve an edge, but this data suggests the more consequential shift is happening upstream, in where researchers choose to start their careers rather than whether established ones leave. For US-based AI investors and founders, the finding is direct evidence that competitive intelligence on Chinese labs' research output deserves at least as much attention as competitive intelligence on other US labs.