AI talent flows are the movements of AI researchers between the country where they complete their education and the country where they go on to work. In U.S.–China AI competition debates the balance of elite researchers is treated as an input to national capability alongside compute, chips and energy, and the flows between the two countries are shaped by immigration and visa policy on one side and by travel and employment restrictions on the other (US-China AI Competition: Different Races, Different Metrics).
Measurement
The most cited longitudinal measure is the Global AI Talent Tracker begun by MacroPolo and continued in 2026 by Carnegie China as AI Talent v3.0. It traces authors of papers accepted at the NeurIPS conference through three career stages: undergraduate institution (the proxy for country of origin), graduate training, and workplace one year after the conference. Carnegie China's 2026 edition, by Damien Ma and Binyi Yang, used the 10,280 of 25,677 unique NeurIPS 2025 authors (40%) with complete career histories on OpenReview (Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026)).
| Measure (NeurIPS cohort) | 2022 | 2025 | Source |
|---|---|---|---|
| Share working in China | 27.1% | 40.6% | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
| Share working in the United States | 46.4% | 34.2% | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
| Share with Chinese undergraduate origin | 46.3% | 57.4% | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
| Share with U.S. undergraduate origin | 19.8% | 13.3% | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
| Chinese-origin researchers working in China | 56.6% | 68.7% | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
| Chinese-origin researchers working in the United States | 31.8% | 21.2% | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
| Chinese-origin in the U.S. per U.S.-origin in China | 46 : 1 | 30 : 1 | Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) |
For the 2025 cohort the study records the United States with the largest net gain (+2,145 researchers: 2,300 arrived, 155 left) and China with the largest net loss (−1,729: 118 arrived, 1,847 left), comparing origin with workplace rather than tracking individual migration. Retention of each country's own graduates was highest in the United States (88.7% of 1,369), South Korea (77.2% of 591) and China (68.7% of 5,903), with Europe at 60.1% of 662 and India at 10.0% of 411. The authors state that the complete-history subset is not random, that broader checks preserve the U.S.–China findings, and that Europe's direction of change is sample-sensitive (Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026)).
A different measure of U.S. dependence on foreign-born talent is cited by Ben Buchanan and Tantum Collins, who state that 70 percent of top U.S.-based AI researchers were born abroad, that 65 percent of leading U.S.-based AI companies as ranked by Forbes have at least one immigrant co-founder, and that before the second Trump presidency 70 percent of students in American AI graduate degrees came from abroad.
Policy levers
United States
U.S. policy has moved in both directions. Executive Order 14110 (October 2023) streamlined visas for AI talent and expanded Schedule A occupations under § 5 before its rescission in January 2025 (Executive Order 14110 (RESCINDED)). Buchanan and Collins write that the 2025 increase of the H-1B fee to $100,000 and visa enforcement against existing students produced a preliminary 30–40% reduction in international AI enrollment, a figure they attribute to NAFSA, and argue that U.S. AI leadership ends if those students go to China (The AI Grand Bargain — Ben Buchanan and Tantum Collins (Foreign Affairs, October 2025)). The Carnegie China study names "tighter visa restrictions," particularly for Chinese graduate students in STEM fields, as one of two likely reasons more Chinese-origin researchers now stay in China, citing the June 2020 presidential proclamation suspending entry of certain students and researchers from the People's Republic of China; the other is easier job-finding in China's AI industry (Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026)).
In Congress, the House Select Committee on the CCP under John Moolenaar has pressed on U.S. AI talent moving to or training Chinese labs, favoring tighter visa controls combined with outbound-investment review, against an industry position, associated with Marc Andreessen, that treats open immigration as the priority (John Moolenaar). The Institute for Progress treats visa and green-card reform as a technology-competition instrument. CSET studies the movement of AI researchers, including Chinese nationals in U.S. labs, and the associated industrial-espionage risk.
China
China has restricted outbound movement of its own AI personnel. Reporting on May 27, 2026 described an expansion of existing travel curbs to top AI researchers, startup founders and private-company executives, some of whom must seek government approval before traveling abroad; it built on March 2025 advisories that top AI founders avoid travel to the United States and followed the barring of Manus's two co-founders from leaving the country during a review of Meta's acquisition of the company (Source: techcrunch.com). See Manus (Butterfly Effect).
Debates and positions
Readings of the same data differ in emphasis. Carnegie China describes the United States as remaining "a magnet for global AI talent, particularly Chinese talent," while describing China as having "established a commanding lead as the largest originating source of elite AI talent globally" and as overtaking the United States as a work location (Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026)). Buchanan and Collins frame foreign-born talent as a condition of continued U.S. leadership (The AI Grand Bargain — Ben Buchanan and Tantum Collins (Foreign Affairs, October 2025)). The Moolenaar committee's position treats the same flows partly as a leakage and security risk (John Moolenaar).
Outside the U.S.–China pair, the Carnegie China study records gains for South Korea and Singapore, whose KAIST, National University of Singapore and Nanyang Technological University rose in its institutional ranking, and declines on most measures for Europe and India (Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026)).
Relationships
- supports: US-China AI Competition: Different Races, Different Metrics — talent as a measure of the competition
- depends-on: Who's Ahead in the Global AI Talent Race? (Ma and Yang, Carnegie China, 2026) — principal current dataset
- related: The AI Grand Bargain — Ben Buchanan and Tantum Collins (Foreign Affairs, October 2025) — immigration as part of a proposed industry–state bargain
- related: MacroPolo, Carnegie Endowment for International Peace, Center for Security and Emerging Technology (CSET), Institute for Progress (IFP), John Moolenaar
- related: Executive Order 14110 (RESCINDED) — § 5 visa streamlining (rescinded)
- related: AI Race Dynamics, AI Sovereignty — talent pipelines as a component of national AI capacity