China Now Hosts More Elite AI Researchers Than the US, Carnegie Study Finds

The numbers come from a think tank inside the Carnegie Endowment for International Peace. Its authors, Damien Ma and Binyi Yang, looked at everyone who had a paper accepted at NeurIPS 2025, the field's most selective conference, and traced more than 10,000 careers from first degree to current employer.

Three years ago the picture was reversed. In 2022, 46% of the group worked in the US and 27% in China, according to the study.

The pipeline has moved too. In 2025, 57% of the tracked researchers took their first degree in China, up from 46% in 2022. More of them are now staying there to work. The study points to job openings across China's AI industry and to tighter US visa rules, especially for Chinese graduate students in science. In Washington, the Trump administration has proposed a $103,265 fee on new H-1B visas.

Where the researchers sit has changed as well. Peking University and Tsinghua University now rank first and second as workplaces, with South Korea's KAIST third. Google held first place in 2022, when the study was last run on comparable data. The National University of Singapore and Nanyang Technological University both moved into the top 20.

The data came from OpenReview, the platform NeurIPS uses for submissions, where researchers report their own career histories. The authors said AI coding tools helped process the data, with human checks on the results.

Europe slips, the US still imports talent

Europe lost ground in the same period. ETH Zurich and the University of Oxford dropped out of the top rankings, according to the study, which also found that Europe kept 60% of its own talent against 89% for the US.

The US still draws researchers from abroad. Among AI researchers working in the US, the share holding a Chinese undergraduate degree rose by four percentage points. The study calls the researchers it tracks the "Navy Seals" of the field. Its sample covered 25,677 authors on 5,823 papers.

Meanwhile, Chinese AI and chip firms are handing out shares to keep their engineers. Last week, Xi Jinping told Donald Trump that the two countries can jointly prevent AI misuse.

None of this is happening in a vacuum. The same week the Carnegie numbers landed, other reporting showed where AI money and AI blame are already moving.

AI costs, and AI care

On 26 September, TechCrunch reported on an analysis by the Blue Cross Blue Shield Association which found that hospitals' use of AI tools as they submit insurance claims led to an additional $942 million in healthcare spending over a two-year period. The analysis described "a sharp increase in patients being documented as having complex conditions" but argued there is a "clear disconnect between [medical] coding and treatment," with "no evidence of corresponding change in care delivered."

The New York Times, cited by TechCrunch, called the analysis the latest sign that AI is contributing to rising healthcare costs. Dr Shiv Rao, founder of the AI startup Abridge, acknowledged the use of AI could lead to "a horrible dystopic future nobody wants to live in," with "bots fighting bots, agents fighting agents," though he said it might also cut costs. The BCBSA's senior vice president Luke Chalker resisted the framing of a battle, saying, "It's not a war. It's a completely one-sided blood bath."

China's health tech push runs on a different logic. Writing in Rest of World on 24 September, physician and consultant Ruby Wang said that in the UK and the US, tech is "always an add-on in the health system," while in China the digital rails are already in place and society treats them as normal. She noted that more clinical trials are conducted in China than in the US, and that China now accounts for half of global drug licensing deals. This year, she said, its share of global drug out-licensing deals could reach 60%.

Wang also flagged the limits. China is strongest in oncology and less mature in immunology, and its biotech firms remain weak at commercialising drugs abroad, which is why out-licensing to partners such as Pfizer is so common. She described US restrictions including the Biosecure Act, the COINS Act and the BINSA bill as a risk to that flow.

For remote clinics and field teams, the connection problem is older and more basic. IEEE Spectrum reported on 21 September that Rhizomatica, a Philadelphia nonprofit, has open-sourced a digital shortwave radio set called HERMES, which sends files and encrypted data over the 3 to 30 megahertz band using roughly 20 watts of power. Founder Peter Bloom told Spectrum the system uses the ionosphere "as our satellite" and can reliably link two radios 400 to 600 kilometres apart, with no satellites or relay towers. He said the point is data, not just voice: manifests, parts lists and telemedicine requests, sent as files instead of read out over the air.