San FranciscoUS AI’s competitive edge runs through immigration offices.
Chinese-born researchers anchor the frontier labs; their visa and travel status has become a live foreign-policy variable in the rivalry with China.
No federal policy yet coordinates AI talent immigration with export controls or technology-competition strategy.
Sources: Rest of World
THE PIPELINE
Chinese nationals earn roughly a third of US AI PhDs, and a majority stay after graduation. The MacroPolo talent tracker found that of top-tier AI researchers working in the US, about 38% did their undergraduate degree in China — more than the US-born share.
THE VISA CHOKEPOINT
Most foreign researchers enter on F-1 student visas, transition to OPT (Optional Practical Training) for up to 3 years in STEM, then compete for H-1B work visas in an annual lottery capped at 85,000 slots. A PhD with a Google offer can still lose the lottery and be forced to leave — a structural fragility no other tech-leading country imposes on itself.
THE PROXIMITY RULE
Frontier AI research happens in a handful of clusters — the Bay Area, Seattle, a few labs in New York and Boston, and parallel hubs in Beijing, Shanghai, and Hangzhou. Tacit knowledge — what works, what doesn't, what the next bottleneck will be — moves at the speed of in-person conversation, not papers.
THE CHINA INITIATIVE PRECEDENT
The DOJ's 2018-2022 China Initiative prosecuted academics for grant-disclosure violations on China ties; most cases collapsed but the chilling effect was measurable. Surveys found a sharp rise in Chinese-born scientists in the US considering departure, and applications from China to US graduate programs softened for the first time in two decades.
THE EXPORT-CONTROL ASYMMETRY
Washington restricts chips, software, and equipment leaving the US for China, but imposes no symmetric framework on talent flows. A researcher trained at a US frontier lab can return to a Chinese lab carrying methods that no export license would have permitted on a USB drive. The know-how walks out legally because it lives in a person, not a file.
THE HISTORICAL RHYME
The US absorbed German physicists in the 1930s and Soviet mathematicians in the 1990s, and each wave anchored a generation of dominance — Manhattan Project, then quant finance and early ML. Every prior tech-leadership era was built on talent the rival country failed to retain.