China now produces more artificial intelligence (AI) research than the US, UK and EU combined — and has become their top research collaborator at the same time.
Western scientists are turning to China more than to each other, while Chinese researchers rely less and less on foreign partnerships. Africa, meanwhile, does not feature at all in the global AI research ecosystem.
These findings come from a new report by London-based research firm Digital Science, which analysed global AI publishing and collaboration data from a comprehensive research information platform. Titled “DeepSeek and the New Geopolitics of AI”, the report details China’s dominant position not only in the volume of AI research, but also in the influence and centrality of its scientific networks.
“China has become the pre-eminent world power in AI research,” writes Daniel Hook, CEO of Digital Science and author of the report. “Not only by research volume, but also by citation attention and influence, rapidly increasing its lead on the rest of the world over the past seven years.”
The UK is much more dependent on Chinese collaboration than it is on US collaboration in AI.
— Daniel Hook, CEO of Digital Science
In 2024 China generated as many AI publications as the US, UK and EU combined. Its papers received more than 40% of all citations to AI research globally that year. By comparison, the US and EU each drew about 10%, while the UK drew just over 2%.
Hook presents AI research as a strategic asset: “Whomever controls the best AI will hold a competitive advantage in a variety of fields,” he writes. “One critical advantage will be an acceleration in research capacity — not just AI research itself but a general acceleration across research more generally.”
China has also emerged as the most connected country in AI science. Researchers in the UK now co-author more AI papers with China than with the US or EU. “This means that the UK is much more dependent on Chinese collaboration than it is on US collaboration in AI,” says Hook.
The EU’s pattern has followed the same trajectory. “China has grown impressively and just recently overtaken both the UK and the US as the major collaborator in AI with the EU-27,” he writes, referring to the 27 member countries of the European Union.
Even in the US, where political efforts have attempted to curb collaboration, Chinese researchers remain the most frequent co-authors in AI.
These partnerships have flowed overwhelmingly towards China, rather than from it. Just 4% of China’s AI research in 2024 included co-authors from the US, UK or EU. In contrast, 14% of US AI publications involved Chinese collaborators, compared to 8% in the EU and over 25% in the UK.
“This paints a clear and undeniable picture that China is the prime connector for AI research,” Hook writes. That role is backed by human capital. China’s AI workforce includes about 30,000 active researchers, compared to 20,000 in the EU and 10,000 in the US.
“China’s combined PhD and postdoc AI populations are twice the size of the US’s total AI population,” says Hook.
This younger cohort is also laying the foundation for sustained innovation. “Given a young, vibrant and highly educated AI-literate workforce, we should anticipate a wave of innovation from China along the same lines as DeepSeek,” he writes.
DeepSeek, the open-source large language model launched in early 2025, was developed using domestic systems rather than US-made Nvidia chips. Hook describes it as a demonstration of capability rather than dependence.
“DeepSeek’s approach clearly demonstrates that China’s AI capability is impressive,” he writes. “It was released under MIT licence ... ostensibly with the intent to level the competitive playing field.”
A parallel analysis by market intelligence firm IDTechEx reinforces this shift towards independence. In a report titled “US Export Controls on AI Chips Boost Domestic Innovation in China”, the firm outlines how trade restrictions on advanced GPUs — the graphics processing chips that are ideal for AI — have spurred development of domestic alternatives.
“Rather than cutting China off from the global AI race,” says IDTechEx technology analyst Jameel Rogers, “export controls have encouraged local industry to innovate in compute architecture, efficiency and model design.”
DeepSeek is cited as one of the most prominent results of this trend. The report frames it as a technical response to sanctions: achieving competitive performance with hardware optimised for energy efficiency rather than brute-force scaling. Rogers concludes that China’s AI ecosystem is “adapting rapidly” and now “advancing along a parallel track” to the US, with an emphasis on “domain-specific models and collaborative deployment across institutions”.
This means innovation in China is not confined to the lab. Patent filings linked to AI research show a steep lead, according to Digital Science. “China’s growth will speed up to an even more impressive level and will continue to grow its advantage in innovation,” says Hook.
Geographic spread adds another layer. In 2024, 156 research organisations across China each produced more than 50 AI publications. These institutions are spread across Beijing, Shanghai, Guangzhou, Nanjing and beyond.
“China has amassed significant volume and density of AI capability in centres all over the country. AI research is taking place at scale across the whole country.”
In contrast, the EU had 54 such organisations, the US had 37, and the UK had 19.
No African country features in the top 12 most influential nations in AI. Even the next eight — including Turkey, Saudi Arabia and Singapore — do not include a single African state. According to the report, no African institution produced more than 50 AI publications in 2024, which was the threshold for inclusion in the geographic breakdown.
South Africa, Kenya, Nigeria and Egypt, often cited as technology leaders in the region, do not register in the core data. Hook’s model for success in AI development places three conditions on national progress: talent, infrastructure and translation of research into economic and social outcomes.
Hook says governments seeking to lead in AI need to: attract or develop and then retain talent; invest in infrastructures to support the development and deployment of AI; and support the generation of routes to translate research into the many different types of benefit.
• Goldstuck is CEO of World Wide Worx and author of The Hitchhiker’s Guide to AI.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.