Traveling just a couple of hours west of Beijing by train reveals the expansive grasslands and ancient volcanic formations of Inner Mongolia. This dry, extensive region has historically been the center of sheep farming and coal extraction in China. However, in recent years, it has emerged as the leading location for AI data center development in the country.
Ulanqab, a city within Inner Mongolia that hosts approximately 1.5 million residents, has seen nearly 100 data centers either launched or under construction since 2016. According to a research report from Goldman Sachs published last week, Chinese enterprises plan to establish operations in the city with an estimated combined capacity of 12.5 gigawatts, over 70% of which has been announced in the last year alone, positioning Ulanqab as one of the fastest-growing computational hubs in Asia. For comparison, OpenAI's ambitious $500 billion Stargate Project is projected to achieve only 10 gigawatts of capacity upon completion.
The influx of Chinese companies to Ulanqab is driven by several factors. The city's high elevation on the Inner Mongolian Plateau coupled with its long, cold winters means that data centers can maintain cooler temperatures with less energy expenditure. Additionally, its proximity to Beijing allows for efficient data transmission with minimal latency. However, the most compelling reason is cost; electricity rates in Inner Mongolia are among the lowest in China, buoyed by the rapid expansion of wind and solar power alongside a rich coal supply.
An intriguing aspect of this data center boom is who is investing in these infrastructures. For the first time, prominent Chinese AI companies are committing significant resources to develop their own facilities instead of relying on external cloud services. Notable players such as DeepSeek, ByteDance, Alibaba, and Xiaohongshu are reportedly constructing substantial AI data centers in Ulanqab. Historically, Chinese AI firms have invested less in physical infrastructure compared to their American counterparts, despite creating several popular and capable AI models. The developments in Ulanqab indicate a shift in this trend as these companies begin to catch up.
However, a pressing challenge looms: the region's water scarcity. Ulanqab experiences a climate similar to Denver, receiving only about 14 inches of rainfall annually. The local government is already grappling with water supply issues to accommodate residents, even before many data center projects are operational. Recently, the municipal water company had to temporarily shut down several waterworks for seven hours nightly to manage peak usage. While the new data centers will require less water for cooling during winter months—based on local weather data showing that additional water is only necessary for two months each year—the increased infrastructure could still significantly strain the region's environmental resources.
The trend towards establishing data centers in Inner Mongolia is not new, having gained traction for over a decade. Huawei initiated its first Ulanqab center in 2016, with Apple soon following suit three years later. In 2021, the region was designated as a central hub for a national initiative known as "Eastern Data, Western Compute" aimed at constructing data centers in China's western territories.
Nevertheless, these remote data centers faced high latency when transferring data to users on the eastern seaboard, complicating their role initially to backup storage solutions. The advent of AI in 2022 transformed their utility; the limited need for real-time processing during lengthy AI model training runs meant that latency became less of a concern. Ulanqab, relatively speaking, is more accessible than many West-based data centers, located closer to major cities like Beijing. Enhanced connectivity through two dedicated fiber optic cables built in 2017 and 2019 has effectively cut average latency to under five milliseconds, making it suitable for real-time data applications such as AI inference.
