Of the 4,622 researchers who published at NeurIPS 2024, 59% list a US institution as their current affiliation. Only 24% received their undergraduate education in the United States. The 35-percentage-point gap — representing over 1,600 researchers — is almost entirely filled by immigrants, primarily from China and India.
This isn't a new pattern, but its scale has grown. America's AI dominance is not primarily a story of superior domestic education or inherent innovative culture. It's a story of visa policy, quality of life, and compensation — the factors that make talented people move.
The Numbers Behind the Narrative
of top AI researchers work in the US
were educated in the US
of China-educated researchers now work in US
of India-educated researchers now work in US
The US is a net importer of AI talent from every major producing country except the UK (which has a roughly balanced flow). China alone accounts for 1,264 researchers who were educated domestically but now work at US institutions — the largest single talent flow in the global AI ecosystem.
Where America's AI Talent Actually Comes From
The composition of US-based AI researchers has shifted dramatically over the past decade. In 2010, approximately 40% of AI researchers at top US institutions were foreign-born. By 2024, that figure exceeds 65%. The shift is concentrated in machine learning and deep learning — the subfields driving current AI advances.
China is the largest source by volume. Among China-educated AI researchers, 72% now work in the United States. The next largest sources are India (80% US-bound), South Korea (45% US-bound), and Europe (40% US-bound). The US brain-drains every major AI-producing region.
Top Producing Institutions → US
The Visa Bottleneck
The US immigration system was not designed for an economy where specialized technical talent is the limiting factor. The H-1B visa lottery rejects the majority of applications through random selection rather than merit. The green card backlog for Indian and Chinese nationals exceeds a decade. PhD graduates from US universities routinely leave for Canada, the UK, or their home countries because they cannot secure work authorization.
Recent policy changes have added friction. Increased scrutiny of Chinese STEM students, restrictions on research collaborations, and general uncertainty about immigration policy have affected application volumes. Chinese student enrollment in US STEM graduate programs declined 8% between 2019 and 2023 — modest in absolute terms, but significant given that this is the largest feeder population for US AI research.
What Happens If the Pipeline Narrows
If foreign-born AI talent stops coming to the United States at current rates, the effects would compound over time. In the short term (2-3 years), US institutions would face talent shortages in specific subfields. In the medium term (5-7 years), the research output gap between the US and competitors would narrow. In the long term (10+ years), US AI leadership would be at risk.
The competitors are not passive. Canada has implemented fast-track immigration for AI talent. The UK launched the Global Talent visa. Singapore and the UAE are offering generous packages to attract researchers. These countries lack the US ecosystem advantages (VC funding, tech giants, research universities) but can partially compensate with immigration policy.
The Core Insight
America's AI advantage is not homegrown talent — it's imported talent. Any policy that restricts immigration from China directly weakens US AI capacity. The question is whether policymakers understand that the talent pipeline is the competitive advantage, not a secondary consideration.
Policy Recommendations
- Staple a green card to STEM PhDs. Graduates of US doctoral programs in AI-relevant fields should receive automatic work authorization and a clear path to permanent residency.
- Merit-based allocation for H-1B. Replace the lottery system with wage-based or skills-based allocation that prioritizes high-value talent.
- Separate research from espionage concerns.Security screening for individuals with specific risk factors should not translate into blanket restrictions on entire populations.
- Domestic investment is necessary but not sufficient.Increased funding for US AI education will help at the margin but cannot replace the volume of talent currently imported.