China is fundamentally recalibrating its approach to artificial intelligence talent, shifting its focus from mass training programs to aggressively recruiting top-tier researchers currently based abroad. While China successfully trains approximately half of the world’s AI talent at the undergraduate and master’s levels, a stark retention problem persists at the highest echelons of the industry. The most brilliant minds—those capable of driving foundational breakthroughs rather than incremental improvements—are overwhelmingly choosing to build their careers in the United States.
According to a comprehensive 2025 study by the Carnegie Endowment for International Peace, which updated the influential Global AI Talent Tracker dataset, the numbers paint a challenging picture for Beijing’s technological ambitions. Of the 100 leading Chinese-origin AI researchers identified in the original 2019 cohort, 87 currently work at institutions or corporations in the United States. Only 10 have returned to China, with the remaining three affiliated with institutions in other countries. This brain drain is particularly evident in elite corporate laboratories; for instance, half of the initial “AI super talent” cohort at Meta’s Superintelligence Labs possessed Chinese educational backgrounds.
The disparity is driven heavily by compensation and research environments. The median AI engineer salary in the US has reached approximately $260,000, compared to roughly $93,000 in China. However, financial incentives are only part of the equation. In recent years, a number of high-profile, Chinese-born AI researchers have made a splash when they decided to leave the United States and return to China. The reasons why any individual researcher returns home are varied and often personal, but the recent geopolitical tumult has created new obstacles and pressures for Chinese researchers who want to remain in the United States.
Beginning in 2018, a series of actual and proposed restrictions on student visas—including discussions of an all-out ban on Chinese students—left many Chinese researchers in limbo. Chinese applicants faced long delays in processing visa renewals, making them uncertain whether they would be able to finish their degrees and remain in the United States to work. Many who did stay said they felt a cloud of suspicion cast on their work by U.S.-China technological tensions and accusations of industrial espionage. A series of high-profile indictments of Chinese researchers in the United States sent chills through these communities, even though many of those cases collapsed upon further investigation. In a 2021 survey of university researchers who self-identify as Chinese, 42 percent reported feeling racially profiled by the U.S. government.
Despite these pressures, the data on Chinese-origin AI researchers showcases the enduring attractiveness of the United States as a place to work at the frontiers of AI. Historically, Chinese researchers who came to the United States for their PhDs have had very high stay-rates, with around 90 percent remaining in the country long term. To counter this enduring trend, Beijing is launching targeted pilot programs, particularly in the Guangdong-Hong Kong-Macau Greater Bay Area, designed to build “regionally embedded talent ecosystems.”
Recognizing that financial incentives alone are insufficient to lure back established researchers who already command top-tier salaries in Silicon Valley, these new initiatives aim to address the non-financial priorities of elite researchers. The focus is shifting toward offering unprecedented research autonomy, massive guaranteed funding blocks without the typical bureaucratic hurdles, and the potential for high-impact work leading national-level projects. As Chinese physicist Wang Yifang noted, the country has an abundance of ordinary researchers, but the critical bottleneck lies in securing the top-tier visionaries necessary to drive foundational breakthroughs.
The Greater Bay Area pilot scheme, recently extended until November 2026, represents a test case for this new approach. It attempts to create micro-environments that mimic the academic freedom and resource availability of top Western institutions, insulated somewhat from broader domestic constraints. Whether these bespoke ecosystems can overcome the gravitational pull of the established US AI hubs remains one of the most consequential questions in the ongoing technological competition between the two superpowers.
The stakes for this talent competition could not be higher. As AI models scale in complexity, the difference between a competent engineer and a visionary researcher becomes exponential. The US has historically benefited from an asymmetric advantage: its ability to attract and retain the best minds from its geopolitical rivals. If Beijing’s new targeted recruitment strategies succeed in reversing this flow, even marginally at the very top of the talent pyramid, it could significantly accelerate China’s timeline for achieving parity in foundational AI research.
This strategic pivot also reflects a growing recognition within Beijing that the traditional state-directed research model may be ill-suited for the rapid, iterative nature of modern AI development. By attempting to replicate the conditions that make Silicon Valley so attractive—high autonomy, massive resources, and a tolerance for failure—China is implicitly acknowledging the limitations of its current system. The success or failure of these new talent ecosystems will serve as a critical leading indicator for the future trajectory of the global AI race.