ULANQAB, Inner Mongolia — Potatoes are still being grown in the fields in the grasslands of Inner Mongolia. Yet another kind of infrastructure is fast emerging nearby—big data centers that are built to provide the computing power required to build and run artificial-intelligence systems.
Once a major potato-producing hub, Ulanqab has become an important node in China’s booming data center industry. WIRED quoted recent research that showed some 100 data center projects have been opened or begun construction in the city since 2016, with planned capacity estimated at 12.5 gigawatts.
The location offers a number of advantages for data center operators, including relatively cheap electricity, a cool climate that reduces cooling requirements, available land, and proximity to Beijing. The rapid growth has also put Ulanqab in the broader push in China to build up domestic computing capacity.
Cranes move materials overhead as workers in hard hats install equipment and build out new facilities at construction sites around the city. A few projects are being built in multiple phases, suggesting that demand for computing power will continue.
The expansion comes as artificial intelligence has become a central component of the technological and strategic rivalry between China and the United States.
The Chinese government has made broader use of AI a national priority. Beijing says under its “AI Plus” initiative, the penetration rate of next-generation intelligent terminals and AI agents should exceed 70% by 2027 and 90% by 2030. The policy also calls for stronger computing infrastructure, AI model development, data resources, and talent.
That goal is more specific than a goal of putting AI into “90% of industries and society,” as the original version of this post stated.
China is also trying to grow its base of AI researchers and engineers. Some Chinese researchers who studied or worked abroad have returned to the country, attracted in part by rising investment in AI and government-backed research programs.
Jack Zhang, of the University of Science and Technology of China, who had studied at Yale University and the University of Cambridge, said he returned because he wanted to apply his AI research to areas such as health and medicine and thought having better resources and institutional support would help him do that work.
The competition has been heightened by Washington’s restrictions on China’s access to some advanced semiconductor technology used in AI development. At the same time, Chinese companies have tried to improve their models and hardware despite those constraints.
The issue is also expected to be a major topic of discussion between President Donald Trump and his Chinese counterpart Xi Jinping in Washington, where the two leaders are also expected to tackle broader US-China tensions.
During recent discussions between US and Chinese officials, a mechanism was proposed through which both sides would be able to inform each other of major incidents related to AI and establish channels of communication in case of potential crises.
AI competition is also shaping the debate in the United States. Trump has argued the US should continue to race ahead in AI rather than slow the development down over safety concerns. Meanwhile, Anthropic CEO Dario Amodei has advocated for maintaining restrictions on the export of advanced AI chips and equipment to China, warning about security ramifications if China gains an advantage in advanced AI.
In some quarters of the AI world, Beijing has promoted a different approach, with more emphasis on open-source development. Chinese companies have released models that external developers are able to access and adapt some of their technology, fostering a wider ecosystem around Chinese AI.
But it would be misleading to suggest that China’s AI industry is all open source. Chinese companies, like US companies, use a mixture of open and closed development approaches.
One of China’s big problems is the supply of high-end computing hardware. US export controls have blocked China’s access to some of the world’s most advanced AI chips, but the restrictions have not deterred Chinese companies from developing alternative hardware and making their models more efficient.
“Recent developments show that progress. The company has announced a new AI chip and plans for a much larger AI model, while other Chinese companies are continuing to invest heavily in domestic computing infrastructure.
The sheer scale of construction in places like Ulanqab shows that China’s AI strategy is more than just about individual models. It is also a bet on the physical infrastructure—data centers, electricity, chips, networks, and skilled workers—that will be needed to support the technology at a national scale.
That infrastructure will be central to the wider US-China competition as both governments try to balance technological development with security concerns and the risks posed by ever more capable artificial intelligence systems.
