New Study: US Gained a Net 2,145 AI Researchers in 2025, While China Lost 1,729
A new Carnegie Endowment study shows that US AI leadership rests on a one-sided inflow of elite Chinese researchers — 30 moved to the US for every one who went the other way in 2025. But the data do not capture recent visa restrictions, and China's own ability to retain researchers is growing fast.
The Headline Number: A One-Way Talent Flow
For every 30 Chinese researchers working on artificial intelligence in the US in 2025, only one researcher in China made the reverse move. That is the core finding of a study from the Carnegie Endowment for International Peace, led by Damien Ma and covered by The New York Times on October 6, 2026.
Measured as net flow, the picture became even starker: the US gained a net addition of 2,145 AI researchers in 2025, while China recorded a net loss of 1,729, according to the report as cited in John Tamny's Forbes commentary via Yahoo and relayed by BigGo Finance.
The numbers land amid a live political tension. Even as the Trump administration tightens visa rules, the study shows that American AI research depends heavily on precisely the researchers that visa policy makes harder to recruit. The study was published and covered the same day — October 6, 2026 — by The New York Times, Newsmax, and Forbes via Yahoo, so the tension between talent dependence and visa policy is immediately relevant.
How the Study Measured Talent Flows
The methodology is simple in concept: Ma and his team went through research papers accepted at NeurIPS, one of the most prestigious conferences in machine learning and computational neuroscience, and mapped where the authors did their undergraduate education, took graduate degrees, and ended up working. In this way, the researchers' movements between countries and institutions were tracked.
The method thus captures the elite tier that publishes at the top level — not the entire AI workforce. Conclusions from the numbers apply to conference authors and cannot readily be generalized to the AI industry or to research environments that do not publish there.
The method also has one significant blind spot: the data do not show when researchers moved. Many of the Chinese researchers counted in the US may have arrived before the administration's latest restrictions took effect, Newsmax notes in its coverage. In other words, the 2025 figures may largely reflect decisions made years ago, before the most recent visa tightening. That means the numbers cannot be used to determine whether visa policy has already had an effect — only what the situation looked like when the researchers made their choices.
The Trends Beneath the Headline Number
Several sub-findings point in the same direction:
- Undergraduate talent: The share of the world's top AI undergraduates of Chinese origin rose from 46 percent in 2022 to 57 percent in 2025, while the share of US-born students fell to 13 percent (BigGo Finance).
- Retention in China: The share of China-born researchers who end up working in China rose from 57 percent in 2022 to 69 percent in 2025 (Forbes via Yahoo).
- Institutions: In 2022, six of the 30 institutions with the most NeurIPS papers were Chinese. In 2025, that number had doubled to 12, with Peking University and Tsinghua ranked ahead of MIT, Stanford, and Carnegie Mellon (Newsmax).
- Authors: Chinese researchers accounted for 41 percent of published NeurIPS authors, versus 34 percent from the US (Newsmax).
Taken together, this points to a two-way movement: China is producing more of the world's elite AI talent and increasingly managing to keep it — while the US still attracts far more than it loses. The Peking and Tsinghua rankings ahead of the top American universities show that the difference is not only about where talent works, but also where it is trained.
China's Recruitment Dilemma
Paradoxically, China struggles with the opposite problem. According to The New York Times, the country has for two years opened its doors to international researchers, with a dedicated research visa and lucrative research fellowships — without succeeding in luring foreign researchers.
China manages to retain its own talent but not to recruit anyone else's. The US does the reverse: recruiting massively from abroad while its homegrown talent base shrinks. Viewed this way, the two countries' strategies are mirror images — and each has an obvious vulnerability.
Significance — and What Remains Uncertain
Damien Ma emphasized to The New York Times that the share of China-born researchers in the US actually rose by four percentage points compared with three years earlier, despite US–China tensions. This suggests that American pulling power has so far withstood the political friction.
But sociology professor Yingyi Ma of Syracuse University warned against reading the numbers as reassurance. "There is a lag effect — the researchers in this study largely made their decisions years ago about whether to study or work here," she told the paper. "The danger for the US is a slow erosion of its appeal to talent that accumulates over decades. By the time it shows up in the data, it is hard to reverse" (Newsmax).
Several caveats are worth keeping clear. All the numbers here are known via secondary coverage — The New York Times, Newsmax, Forbes via Yahoo, and BigGo Finance — not from the report in its original form, which limits how precisely the methodology can be verified. Moreover, NeurIPS authors represent only a slice of AI research, not the entire workforce. And BigGo Finance's interpretation that the visa tightening is "accelerating a reversed brain drain" is that outlet's analysis — the study's data cannot establish that causal link, precisely because the timing of moves is not visible. The same applies to the shift in Chinese retention from 57 to 69 percent: the data show the trend, but not what drove it.
The honest answer is therefore two-sided. The figures as of 2025 show a US still winning the talent war decisively, with a net inflow that China cannot match and that China's own recruitment efforts have failed to counteract. At the same time, all the trend lines — a higher share of Chinese top-trained talent, better Chinese retention, more strong Chinese institutions — point toward China's position improving over time. Whether the US advantage holds once decisions made under the new visa rules become visible in the data — likely years from now, as the lag effect Yingyi Ma describes suggests — is an open question none of the sources can yet answer. What can be said with confidence is that US AI leadership today is built to a significant degree by researchers whom visa policy now makes harder to bring in — and that the next round of this study will be the first real measure of whether the policy carries a price.