What happened
A new report from Carnegie China, titled 'Who’s ahead in the global AI talent race', indicates that China has surpassed the United States as the primary workplace for elite AI researchers. The study, which analyzed 10,280 authors from the 2025 Conference on Neural Information Processing Systems (NeurIPS), found that China employed 41% of these top-tier researchers in 2025, compared to 34% in the US. This represents a significant reversal from 2022, when the US held 46% of the talent pool against China’s 27%.
The Carnegie China study, authored by Damien Ma and Yang Binyi, utilized a of 10,280 researchers who published at the 2025 NeurIPS conference. By tracking the undergraduate origins and current employment locations of these individuals, the researchers mapped a clear trend of increasing domestic retention within China.
According to the report, 57% of elite AI researchers earned their undergraduate degrees in China, an 11-percentage-point increase from 2022. Simultaneously, the retention rate for these graduates has risen to 69%, up from 57% in 2022. The authors attribute this to a combination of 'push and pull' factors, including competitive compensation packages in China and increased barriers to entry in the US.
The report also highlights that while China has overtaken the US in total employment of elite talent, the US remains the leading destination for researchers who choose to work outside their home country. The share of US-based researchers who earned their undergraduate degrees in China actually rose by 4 percentage points last year, indicating that American institutions continue to attract top-tier Chinese talent despite broader geopolitical tensions.
Why it matters
The shift in talent distribution highlights a fundamental change in the global AI landscape, suggesting that China’s domestic ecosystem—bolstered by firms like DeepSeek and Moonshot AI—is successfully retaining its own graduates. With 69% of Chinese-educated researchers now remaining in the country, the findings suggest that the 'brain drain' to Silicon Valley is slowing. This trend is compounded by geopolitical friction, including US visa restrictions and new Chinese exit controls, which are increasingly segmenting the global AI labor market. As both nations prioritize AI supremacy, the ability to attract and retain elite researchers has become a critical metric of national technological power, potentially influencing the pace of innovation and the development of frontier models in both jurisdictions.
The findings underscore the success of China's strategy to build a self-sufficient AI ecosystem. By fostering high-profile domestic companies, China has created a professional environment that rivals Silicon Valley in terms of both compensation and technical challenge, reducing the necessity for top researchers to seek opportunities abroad.
Geopolitical constraints are playing a significant role in this shift. US visa restrictions targeting researchers with links to institutions associated with China’s 'military-civil fusion' strategy have limited the flow of talent into the US. Conversely, China’s own recent regulations, which restrict the movement of citizens who violate export controls, further complicate the international mobility of high-level technical staff.
This talent concentration is occurring alongside aggressive hardware development. The report notes that Chinese firms are accelerating efforts to build domestic AI infrastructure, such as Alibaba’s recent unveiling of new AI chips and its plans for the Qwen 5 model, which aims to utilize 10 trillion parameters.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
What did neural scaling law research (e.g. Kaplan et al., 2020) observe?
What to watch next
Observers should monitor whether the US can maintain its status as the primary destination for international researchers who do choose to move abroad, as the study noted that the US remains the top destination for mobile talent. Additionally, the impact of tightening export controls on advanced hardware and the reciprocal restrictions on researcher mobility will likely continue to shape the composition of AI labs. Future developments in the 'Qwen' model family and other domestic Chinese projects will serve as indicators of whether this talent retention translates into sustained competitive advantages in model performance and scale.
The long-term impact of these mobility barriers on the quality of research output in both nations remains a key unknown. While China is retaining more talent, the global nature of AI research often relies on cross-border collaboration, which may be hampered by the current climate of scrutiny.
The study provides a snapshot based on 2025 conference data; future iterations of this research will be necessary to determine if this trend represents a permanent structural shift or a temporary fluctuation influenced by current policy cycles.
The practical implication for global AI development is a potential bifurcation in research ecosystems, where the 'talent race' becomes as much about policy and visa regimes as it is about academic prestige or corporate compensation.