China’s AI Data Center Boom Raises Water Risks in Inner Mongolia
Ulanqab in China’s Inner Mongolia has emerged as a computing hub for artificial intelligence, with technology companies including Huawei, Tencent and DeepSeek building infrastructure for training and running large language models. The expansion highlights the physical costs behind China’s AI ambitions: data centers consume large amounts of electricity and require intensive cooling, placing additional pressure on water supplies in a region already vulnerable to scarcity and drought.
A recent report said 89 AI data centers have been established in Ulanqab. Heat generated by their servers is partly removed through water-intensive cooling and evaporation, raising concern that continued expansion could worsen local shortages and drought risks. The report called on operators to disclose water withdrawals, consumption and recycling data, and urged technology companies to share the cost of expanding the water, electricity and utility networks needed to support the computing parks.
All Coverage
1 original reportsThe Backstory
The history behind this eventChina Plans 12.5-GW AI Data Center Hub in Inner Mongolia
Ulanqab, a city in China’s Inner Mongolia region, is emerging as a major artificial-intelligence infrastructure hub, helped by its cool climate, relatively inexpensive electricity and low-latency links to key markets. The buildout illustrates how China is using energy-rich inland areas to support power-intensive model training, as access to computing capacity, chips and reliable electricity becomes increasingly important in the global AI race.
Data centers planned for Ulanqab have a combined capacity of 12.5 gigawatts, a scale the report said exceeds OpenAI’s Stargate project. The figure represents planned capacity rather than facilities already completed or operating, an important distinction when comparing the projects. DeepSeek, ByteDance and Alibaba are among the technology companies investing in the area, turning the remote city into a core base for training large AI models.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.
If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →