Tsinghua Hephaestus Team, known for its work in advanced robotics and AI, competes in the World Humanoid Robot Games in Beijing in August.
Source: VCG/Getty Images
Xi’s University Fuels China AI Boom With More Patents Than Harvard or MIT
Tsinghua University has educated the country’s top science and engineering students for decades. Now, it’s at the forefront of the AI revolution.
By Saritha Rai
November 18, 2025 at 5:00 PM EST
On a crisp afternoon in Beijing, the campus of Tsinghua University hums with the activity of the country’s top students in science and engineering. Badminton courts near the school’s east entrance echo with grunts and shouts. In a sleek new wing housing the Laboratory of Brain and Intelligence, the work is quieter but no less determined. Researchers, amid whiteboards dense with equations and the acrid smell of fresh paint, try to decode the inner workings of the human mind.
Tsinghua is buzzing these days. While the school has long been the nation’s leading institution for science and technology — Stanford University, Massachusetts Institute of Technology and Carnegie Mellon rolled into one, as some describe it — this has been a breakthrough year for China. The success of DeepSeek, the artificial intelligence startup that stunned the tech industry with its innovative large language model, has brought a newfound confidence to these young brainiacs that they have an opportunity to compete with — or beat — the world’s best. The school’s graduates have founded at least four of the country’s top AI startups, so far.
“DeepSeek showed that a Chinese team could lead in the LLM race,” says Yuyang Zhang, a 26-year-old who is working on a PhD in computational biology at Tsinghua.
The sense of ambition is growing on the university’s 1,200-acre campus, where students on Tsinghua’s signature purple-hued bikes glide past historic Qing dynasty buildings and stylish cafes. The school has demonstrated its brainpower for years, leading US News & World Report’s global rankings in subjects such as engineering, artificial intelligence, computer science and chemical engineering.
What’s different now is the alchemical opportunity to turn intellectual accomplishment into gold and glory. Xi Jinping and the Communist Party have called on the private sector to help develop critical technologies — especially artificial intelligence — and they’ve backed that up with tax breaks, subsidies and supportive policies. Not only can founders like DeepSeek’s Liang Wenfeng raise millions in venture capital and build businesses, their photos are splashed across state media alongside Xi’s where they’re hailed as national heroes. Indeed, China’s president is a graduate of the school too.
In all this, the university is having an outsized impact. Beyond the alumni who have founded startups, its graduates play key AI roles at major tech companies, including Alibaba Group Holding Ltd. and ByteDance Ltd. Within Tsinghua’s labs, researchers have developed an AI chip called Accel to compete with those from leader Nvidia Corp., a drug discovery engine called DrugCLIP and a training protocol known as Absolute Zero Reasoner that allows AI models to learn on their own, without any human-provided data.
The school’s professors and students are quietly accumulating massive amounts of intellectual property. They already had more AI research papers among the 100 most cited than any other school at last count, and they receive more patents each year than MIT, Stanford, Princeton and Harvard combined. Tsinghua collected 4,986 AI and machine learning patents between 2005 and the end of 2024, including more than 900 last year, according to the data analytics service LexisNexis. Overall, China now accounts for more than half of all active patent families globally in those fields.
“This is a staggering shift in innovation in less than a decade, and it reflects China’s concerted drive to become an AI superpower,” said Marco Richter, LexisNexis’s Bonn-based senior director of IP Analytics and Strategy.
The university anchors a nationwide education strategy that begins in elementary school, where AI is now taught alongside math and language. This effort results in a much broader tech-savvy workforce than in the US: China graduated 3.57 million students in science, technology, engineering and math (STEM) in 2020, compared with 820,000 in the US, according to the Washington, DC-based think tank Center for Strategic and International Studies. China’s state-backed People’s Daily said last year that the annual number of STEM graduates had reached five million.
Tsinghua Outpaces US Universities in Producing AI Patents Patent publications in machine learning and artificial intelligence each year
China’s education system had long been handicapped by focusing on memorization and standardized testing, the kind of rote learning that stifles critical thinking. But Tsinghua is an example of how this is changing. One of its legendary professors is Andrew Chi-Chih Yao, the country’s only Turing Award winner, who returned to teach in China after years at Princeton, Stanford and MIT. He founded an innovative computer-science program – the Yao class – that became renowned for its rigorous, interdisciplinary approach to teaching. The Laboratory of Brain and Intelligence built on that tradition by exploring the intersection of areas like computational neuroscience, engineering and computer science.
The AI startup Sapient Inc. is the result of this cross-disciplinary approach. It was in the Brain lab in 2023 that two undergraduates, Guan Wang and William Chen, began sketching the outlines of an AI system inspired by the brain’s layered reasoning. Their early experiments grew into Sapient’s Hierarchical Reasoning Model, or HRM, that mimics the brain’s processing — slow, methodical planning paired with split-second reflexive responses. The model has outperformed larger models from OpenAI and Anthropic on reasoning benchmarks and complex Sudoku puzzles.
Wang and Chen, now both 24, see their work as an alternative path to artificial general intelligence, in contrast with the large language models now popular in Silicon Valley. Their strategy is to build a general-purpose AI capable of mastering any task at human or superhuman levels.
“We have a unique technical pathway to AGI,” says Chen, during an interview in the office he retains on campus. “We want to build an AI model architecture that's 10 times better than existing models.”
Note: Top 20% of patents as defined by the Patent Asset Index. Technology relevance score (where 0 is the lowest) scores how often a university's patents are cited by later inventions, indicating the impact of its research. Worldwide average for all patents is a score of 1. Portfolio size indicates the number of patents, including active patent families, held by each university at the end of each year.
Chen, who graduated from high school in Michigan, got accepted to Carnegie Mellon and Georgia Tech, but opted for Tsinghua because of his family’s roots in the country and the opportunity to turn his research into a business. "It was an unusual path for a US high school student,” he says, dressed in black jeans and a black T-shirt.
Rocky Xia leads the university’s startup incubator, making him the point person for helping students turn their ideas into startups. He took a visiting reporter on a tour of Tsinghua X-lab to show how the school supports would-be entrepreneurs. Splashy posters advertise workshops, lectures and bootcamps. One features Tang Jie, co-founder of a hot startup called Z.ai (formerly Zhipu.ai) who got his PhD at Tsinghua and is now professor of computer science there. Another shows Wang Xingxing, who started the country’s leading robotics company, Unitree. Xia, who steers a Shark Tank-like competition for entrepreneurs at the university called the President’s Cup, says the X-lab has spawned about 900 startups since its inception in 2013. “There’s a big focus around AI,” he says, dressed in shorts, a T-shirt and grey-and-white mules.
Jun Liu, a former professor at Harvard and one of its top statisticians, returned to China this year to establish a new Department of Statistics and Data Science at Tsinghua, and is actively recruiting from top US universities. He was born on the university’s campus to faculty parents, and now lives just outside the south gates.
“There’s a lot of enthusiasm for AI and machine learning within government, industry and academic circles,” Liu says, during an interview in his new office. “The draw of AI talent is due to capital, and the Chinese government’s support for scientific research, including in AI and related areas.”
Tsinghua is integrating AI across the university. AI and large language models are now embedded in daily research across disciplines. Competitions to build domain-specific AI agents, once the purview of the computer science elite, are commonplace. In September, Tsinghua launched a new AI computing platform, which gives all students at the school subsidized access so they can experiment with new models. “Every student, regardless of their major, can receive free computing credit to use for any research,” Zhang said.
The US still holds the most influential AI patents and the top performing models. Harvard and MIT, for example, consistently rank ahead of Tsinghua in the influence of their patents. American institutions also produce more notable AI models than China — 40 to 15 in 2024, according to Stanford’s AI Index Report — though Chinese organizations are closing the gap in certain performance benchmarks.
The US may struggle to keep its place as the leading country for AI researchers. China’s share of the world’s elite AI researchers — the top 2% — rose from 10% in 2019 to 26% in 2022, while the US share fell from 35% to 28%, according to Information Technology & Innovation Foundation, a Washington, DC-based think tank. (The data is the most recent available from the foundation.)
Zhang, the computational biologist, thinks Tsinghua graduates are more inclined to remain in China these days.
“Most of my classmates will stay in China,” says Zhang. "I find Tsinghua in its most vibrant state right now.”
— With assistance from Rachel Lavin
bloomberg.com
|
