Ulanqab, a city in Inner Mongolia, as an unlikely center of China’s rapidly expanding AI infrastructure. Nearly 100 data centers have opened or begun construction there since 2016, with planned projects representing about 12.5 gigawatts of capacity—more than the 10 GW target for OpenAI’s Stargate project. Chinese companies including DeepSeek, ByteDance, Alibaba and Xiaohongshu are investing heavily in the area.
Ulanqab is attractive because it combines cheap electricity, cold weather, available land and relatively low network latency to Beijing. Its cold climate reduces cooling costs, while Inner Mongolia has abundant wind, solar and coal power. AI has also changed the economics of locating data centers far from China's population centers: model training can run for months and is less sensitive to latency than consumer-facing applications. New fiber connections have also reduced latency to below five milliseconds, making the region increasingly suitable for AI inference.
The boom represents a significant change for Chinese AI companies, which historically relied more heavily on cloud providers instead of building their own massive computing infrastructure. The government's “Eastern Data, Western Compute” strategy also encourages moving computing capacity toward western regions where land and energy are cheaper and renewable-energy capacity is abundant. Companies are increasingly using these data centers to absorb surplus renewable electricity, although the reality is more complicated: around 37% of Ulanqab's electricity still comes from coal.
The biggest constraint may ultimately be water rather than electricity. Ulanqab receives only about 14 inches of rain annually and already faces pressure on its water supply, while dozens of new data centers could increase demand for cooling. The story therefore illustrates a broader reality of the AI infrastructure race: countries need not only chips and computing power, but also enormous quantities of land, electricity, fiber connectivity and water. China's AI expansion is increasingly becoming an infrastructure story—and Ulanqab shows both the economic opportunity and environmental costs of building that infrastructure at scale.
