China has expanded visas for scientists and provided lucrative research grants, yet still struggles to attract foreign AI researchers.
Research from the Carnegie Endowment shows that in 2025, for every 30 Chinese-origin AI researchers working in the US, only 1 researcher moved in the opposite direction to China.
Notably, the proportion of Chinese-origin researchers working in the US increased by 4 percentage points compared to three years prior, despite US-China tensions.
The US maintains its advantage through its ability to attract international talent from China, India, South Korea, and many other countries.
China has made strong progress in retaining domestic talent as an increasing number of researchers nurtured within China’s education system choose to stay.
Among the 30 institutions publishing the most research at NeurIPS, the number of Chinese establishments increased from 6 in 2022 to 12 in 2025.
Peking University and Tsinghua University top the list, outperforming US institutions such as MIT, Stanford, and Carnegie Mellon.
Chinese authors account for 41% of authors publishing at NeurIPS, higher than the US figure of 34%.
DeepSeek and Moonshot are elevating the professional prestige of China’s AI ecosystem, but Silicon Valley remains more attractive to many international scientists.
📌 Conclusion: The AI talent race reveals two different models: China is growing stronger through its domestic training system and ability to retain local human resources, while the US maintains its advantage through global talent attraction. China achieved 41% of NeurIPS authors and boasts 12 out of the top 30 institutions, but cross-border talent flows remain heavily imbalanced at a 30:1 ratio. Therefore, high research output does not yet equate to the ability to become a global AI talent hub.
