Ulanqab, an Inner Mongolian city of about 1.5 million people, has become a major destination for Chinese AI infrastructure. Nearly 100 data centers have opened or entered construction there since 2016, and announced projects total an estimated 12.5 gigawatts, according to a Goldman Sachs research note. More than 70% of those commitments arrived in the past year.
The city combines relatively cheap electricity, abundant land, cold winters that reduce cooling demand, and a location roughly two hours from Beijing by train. Dedicated fiber links have brought average latency below five milliseconds. Remote facilities once used mainly for backup storage can now support months-long model training, while improved connections also make real-time inference practical.
DeepSeek is reportedly building a large facility in Ulanqab, alongside projects from ByteDance, Alibaba, and Xiaohongshu. Envision has announced a two-gigawatt data center connected to its own clean-power supply. The buildout could absorb surplus wind and solar generation, but the region is not yet fully renewable: research cited in the article says about 37% of Ulanqab’s electricity still comes from coal.
Water presents another constraint. The dry city receives roughly 14 inches of rain annually, and its water company recently shut several waterworks for seven hours nightly to manage peak demand. Cooling needs are seasonal, but the planned scale could still create a significant environmental challenge.