Inside China’s Race to Build AI Data Centres at ‘China Speed’

In Ulanqab, a stretch of Inner Mongolia grassland long known as China’s “potato capital,” farmers now harvest their crops within sight of a new kind of neighbour: sprawling data centres designed to power the country’s artificial intelligence ambitions. Construction here has accelerated even as Donald Trump and Xi Jinping prepare to discuss AI safety when the Chinese leader arrives in Washington this week.

“This is China speed,” one construction worker joked at the site, where cranes and scaffolding dominate the skyline and welding sparks fly late into the day. An electrical engineer on the project said it was only the second of several planned building phases.

Computing power is central to Beijing’s rivalry with Washington over artificial intelligence, and the Chinese Communist Party regards facilities like this one as matters of economic and national security — part of a push to weave AI into 90% of industries and society by 2030. China is erecting hundreds of such centres, serving firms including Huawei, ByteDance and DeepSeek, across thinly populated regions such as Ulanqab, Ningxia, Gansu and Guizhou, where land is cheap, the climate is cooler, and wind and solar capacity is already extensive.

That expansion isn’t free of friction. Much of the regional power grid still runs on coal despite Beijing’s 2030 green-energy target, water remains scarce in Inner Mongolia’s dry terrain, and entire villages have reportedly been relocated — a subject few locals are willing to discuss openly. “This big data has absolutely nothing to do with ordinary people,” one resident said anonymously. “What benefits has it brought us?” Not far away, another villager, giving only his surname Zhang, offered a more upbeat take while tending potatoes near the site: “China is now at the forefront of the world,” he said, adding that his own children had found jobs there.

A talent pipeline redirected homeward

Washington’s tighter immigration rules and rising US-China tension are also reshaping where young researchers choose to build careers. “Fewer students and researchers are heading to the US, and more are returning from the US to China. That is very clear,” said Jack Zhang of the University of Science and Technology of China, who was recruited home after studying at Yale and Cambridge. China now produces around five million science and technology graduates annually, compared with roughly half a million in the US.

Washington, meanwhile, shows little appetite for slowing down. “Whoever wins AI, wins,” Trump said last week, while Anthropic chief executive Dario Amodei has urged keeping US restrictions on advanced chip sales to China, warning that a Chinese lead in the technology would “pose grave danger for the US and the world.” Beijing’s state-run Global Times dismissed such warnings as a “Cold War playbook” aimed at containing China’s rise.

Rather than simply chase Washington, Beijing is betting on a different strategy: encouraging Chinese firms to release cheaper, open-source AI models that rivals such as OpenAI and Anthropic keep proprietary. The approach has let Chinese developers iterate quickly despite restricted access to advanced chips, and Xi has pushed for a larger Chinese role in setting the rules of any future global AI framework — a debate now playing out at this week’s White House summit on AI safety.

Source: BBC News