AI talent flows reverse
Reports say a wave of prominent AI researchers has moved back from the U.S. to China over the past year, driven largely by economics and opportunity. (alltoc.com) Analysts and coverage also frame this as part of a broader AI rivalry that now hinges on talent, infrastructure and industrial policy, not just export controls. (bbc.com)
A growing share of elite artificial intelligence researchers trained in the United States are taking jobs in China, as Chinese universities, labs and start-ups offer bigger roles and faster funding. (technologyreview.com) MacroPolo’s 2024 update, based on researchers who presented or published at the NeurIPS conference in December 2022, found the United States still employed the largest share of top artificial intelligence talent, but more researchers were staying in their home countries instead of moving abroad. Chinese-origin researchers made up 47 percent of that top-tier pool. (technologyreview.com) By 2022, 42 percent of the world’s top artificial intelligence researchers worked in the United States and 28 percent worked in China, down from 59 percent and 11 percent in 2019, according to a January 2025 report from the White House Council of Economic Advisers cited by MIT Technology Review. (technologyreview.com) The shift is showing up in individual moves. Channel NewsAsia reported in October 2025 that researchers including Fu Tianfan, a 32-year-old former tenure-track assistant professor at Rensselaer Polytechnic Institute, had returned to China for posts in Nanjing, Beijing and Shenzhen. (channelnewsasia.com) Fu joined Nanjing University in December 2024 after leaving New York, and told the South China Morning Post that China’s investment in higher education had created “unprecedented opportunities” for younger scientists working on artificial intelligence for drug discovery and materials research. (scmp.com) China has spent years building the pull factors behind those decisions. Nature reported in June 2025 that hundreds of policies across China were designed to lure researchers and new graduates to settle there, and followed in July 2025 with a report that many returnees received high salaries, housing support and senior positions. (nature.com 1) (nature.com 2) That recruitment drive is landing in a research system that is already gaining ground. Nature Index said China’s 2024 share of papers in 145 leading natural- and health-science journals reached 32,122, compared with 22,083 for the United States, widening a lead China first took in 2023. (nature.com) Artificial intelligence competition also depends on chips, data centers and model distribution, not just on star scientists. MIT Technology Review reported in March 2025 that China had poured billions into artificial intelligence data centers, even as some newer facilities sat underused after a speculative building boom. (technologyreview.com) On the software side, Chinese firms have pushed open-weight models, which let outside developers download and modify the system instead of renting access through a closed service. MIT Technology Review reported in February 2026 that Alibaba’s Qwen family had overtaken Meta’s Llama in cumulative Hugging Face downloads, and that Chinese open-source models had surpassed United States models in total downloads in a recent Massachusetts Institute of Technology study. (technologyreview.com) The United States still leads many frontier labs and remains the top workplace for elite artificial intelligence researchers, but the older pattern — train in America, stay in America — is no longer holding as firmly as it did a few years ago. (technologyreview.com)
