The new brain drain: China makes AI brains, America buys them, India loses them
In AI, brain drain means a country loses researchers it has educated when they move abroad to work. It matters because the article describes human intelligence as the most valuable resource in the contest for AI leadership. Countries need researchers who can build advanced systems, and their work can shape economic and geopolitical power. The story names Fei-Fei Li, a Stanford AI pioneer; Kaiming He, whose ResNet architecture transformed computer vision; and Anima Anandkumar, an IIT Madras alum and Caltech-Nvidia collaborator. Anandkumar pioneered Neural Operators, techniques that can learn and simulate complex physical phenomena. Her career in the US illustrates how India can educate researchers who later work elsewhere. The Carnegie study found that roughly 370 of 411 India-educated researchers in its elite cohort went abroad, overwhelmingly to the US. Only about 10% stayed in India. The figures point to a challenge beyond producing talent: India also needs to retain researchers, while the US relies in part on researchers educated abroad.
What does “brain drain” mean in AI, and who are the “AI brains” in this story?
In AI, brain drain means a country loses researchers it has educated when they move abroad to work. It matters because the article describes human intelligence as the most valuable resource in the contest for AI leadership. Countries need researchers who can build advanced systems, and their work can shape economic and geopolitical power.
The story names Fei-Fei Li, a Stanford AI pioneer; Kaiming He, whose ResNet architecture transformed computer vision; and Anima Anandkumar, an IIT Madras alum and Caltech-Nvidia collaborator. Anandkumar pioneered Neural Operators, techniques that can learn and simulate complex physical phenomena. Her career in the US illustrates how India can educate researchers who later work elsewhere.
The Carnegie study found that roughly 370 of 411 India-educated researchers in its elite cohort went abroad, overwhelmingly to the US. Only about 10% stayed in India. The figures point to a challenge beyond producing talent: India also needs to retain researchers, while the US relies in part on researchers educated abroad.
How many leading AI researchers were educated in China, the US, and India—and where do Indian-educated researchers go?
The Carnegie analysis tracks 10,280 researchers whose papers were accepted at NeurIPS 2025. Nearly 60% received their undergraduate education in China, compared with 13% in the US. The article also reports that India accounted for 411 researchers in this elite cohort. These figures show how concentrated the research talent is in China and the US.
Education and current workplace are different measures. China-educated researchers made up nearly 60% of the cohort, but 41% of all researchers were working in China. Among the 3,514 leading researchers working in the US, 35.7% were educated in China and 34.5% in America. Together, those groups represented about 70% of the US research workforce studied.
India retained only about 10% of its 411 India-educated researchers; roughly 370, or 90%, went abroad, overwhelmingly to the US. The article also reports 285 India-educated researchers, or 8.1%, in its comparison of leading researchers. The different figures are reported in separate parts of the study discussion, so they should not be treated as identical measures.
How does the US benefit from attracting researchers educated in other countries?
The US benefits when researchers educated in other countries choose to work there: it gains skilled people whose work contributes to its AI research capacity. The article links AI expertise to economic and geopolitical power, so attracting leading scientists matters to the US effort to preserve technological primacy. This is especially significant because many researchers in the US workforce studied came from elsewhere.
The Carnegie study counted 3,514 leading researchers working in the US. About 35.7% were educated in China and 34.5% in America. Together, those groups made up about 70% of the US elite AI research workforce. The figures show a concrete mechanism: the US draws on people educated in China as well as on those educated at home.
The article describes the US as a talent magnet, but says political suspicion around Chinese access to advanced technology is creating uncertainty. It argues that an unsettled international scientific community could affect the very talent base on which US technological leadership depends. Maintaining an environment that attracts researchers is therefore important to sustaining that leadership.
Why is India’s AI talent gap about both producing fewer advanced researchers and retaining the ones it does produce?
India’s AI talent gap has two parts: it produces fewer advanced researchers and retains too few of those it does produce. Both matter because building a strong AI research base requires a supply of trained people as well as institutions and opportunities that keep them working in the country. The article says India trails China and the US on several measures.
The doctorate figures show the production gap. India awarded around 23,000 science and engineering doctorates in 2021, while China awarded about 53,000 in 2022 and the US 45,000 in 2023. In computer science, India produced 621 PhDs, compared with 2,874 in the US. The retention figures show another weakness: about 370 of 411 India-educated researchers went abroad.
Research impact and innovation measures also reveal a gap. India’s share of global citations in computer science AI research rose to 6% in 2023, while China’s was about 22.5%. China held 70% of granted AI patents, while India’s share was negligible. Building research capacity and keeping researchers could therefore both matter to India’s future AI strength.
What could happen to US AI leadership if immigration rules make international researchers feel unwelcome or insecure?
The US benefits from researchers educated around the world, but the article describes a climate of growing suspicion toward Chinese access to advanced American technology. If immigration policies make international scientists feel unwelcome or insecure, they may be less able or willing to work in the US. That could put pressure on the talent base supporting American AI research.
The story gives the case of Eva Yi Xie, a Chinese-born Princeton doctoral student and MIT graduate. Immigration authorities detained her at Reagan National Airport as she returned from a neuroscience conference. Her lawyers said she had valid work authorization; the Department of Homeland Security disputed her authorization to study. She had not been accused of espionage or any connection to the Nvidia investigation.
The article says Xie’s case unsettled academic circles and illustrates how immigration restrictions can unsettle the international scientific community. It does not claim that the case has already changed US AI leadership. But because the article says that leadership depends on this community, making researchers feel insecure could make it harder to preserve the US talent advantage.
How do PhDs, research citations, patents, and notable AI models help show a country’s strength in AI?
No single number captures a country’s AI strength. The measures in the article look at different stages of research and innovation. PhDs indicate how many advanced researchers a country produces; citations indicate how often other researchers refer to its work. Patents and notable models offer further signs of whether research is being turned into inventions and prominent AI systems.
The figures show how these measures can tell a connected story. India awarded around 23,000 science and engineering doctorates in 2021, but produced 621 computer science PhDs. Its share of global citations in computer science AI research rose from about 3.2% in 2013 to 6% in 2023, while China’s was about 22.5%. China held 70% of granted AI patents in 2023, while India’s share remained negligible.
The Stanford AI Index adds a measure of notable model output: US institutions produced 40 notable AI models in 2024, compared with 15 from China. India produced none that met the index’s threshold. Together, the indicators show differences in talent pipelines, research influence and visible innovation, while each measures a distinct part of AI capacity.
What is a neural network, and how do researchers use it to build AI models?
The article does not define neural networks. In established AI usage, a neural network is a computing system made of connected units that adjusts internal settings as it learns from examples. Researchers use networks to find patterns in data and build models that can perform tasks based on those patterns. This general explanation provides context for the article’s discussion of AI research.
Kaiming He’s ResNet architecture transformed computer vision by making it possible to train much deeper neural networks, according to the article. Anima Anandkumar pioneered Neural Operators, a newer class of AI techniques that can learn and simulate complex physical phenomena. These examples show researchers developing methods that let AI systems handle more complex tasks.
The article presents such breakthroughs as part of the competition for AI leadership. It does not explain the detailed mechanics of training neural networks or give a technical account of these methods. It does show that researchers who develop architectures and techniques are central to shaping AI, and that countries’ ability to produce and attract such researchers matters to their technological position.
This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.
Read more in the JupiteX app
Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.
Or read more news on the web