Chinese AI Firms Build 12.5-Gigawatt Hub in Ulanqab

The remote city of Ulanqab in Inner Mongolia has rapidly transformed into the epicenter of China's artificial intelligence infrastructure boom. According to a recent Goldman Sachs research note, Chinese companies have committed to building projects with a combined estimated capacity of 12.5 gigawatts in the city of 1.5 million people, with over 70 percent of these commitments announced within the past year. To put this scale in perspective, OpenAI's planned 500 billion dollar Stargate Project is designed to reach 10 gigawatts of total capacity. Nearly 100 data centers have been opened or have begun construction in Ulanqab since 2016.

This surge represents a fundamental shift for AI practitioners in China. Rather than renting cloud computing power, major firms like DeepSeek, ByteDance, Alibaba, and Xiaohongshu are investing directly in their own physical infrastructure. The high-elevation plateau offers cold winters that reduce cooling energy, while its proximity to Beijing keeps latency low. Two dedicated fiber optic cables installed in 2017 and 2019 have slashed average latency to under five milliseconds. This allows startups like Moonshot AI and Zhipu AI to utilize the hub not just for long-term model training, but also for real-time inference to serve a growing base of paying domestic users.

However, the rapid expansion faces critical resource constraints. Ulanqab is highly arid, receiving only about 14 inches of rain annually. Last month, the local water utility had to shut off several waterworks for seven hours each night to manage peak demand. While data centers there only require extra cooling water during two months of the year, the environmental strain remains high. On the energy front, the government hopes the hub will absorb excess renewable energy, prompting wind turbine manufacturer Envision to announce a 2 gigawatt AI data center powered by its own clean supply. Yet, coal still generates roughly 37 percent of Ulanqab's electricity, meaning practitioners must balance their scaling needs against local environmental impacts and fossil fuel reliance.