A city better known for grasslands, cinder cones, sheep farming, and coal mining is becoming one of the most important places in China's AI buildout. Ulanqab, in Inner Mongolia, is now drawing major data center commitments as Chinese technology companies push deeper into artificial intelligence infrastructure.
The rush is not accidental. Ulanqab offers a rare mix of low-cost power, cold weather, and workable proximity to Beijing. But the same factors that make the city attractive also expose the tradeoffs behind the AI boom: more computing capacity means more pressure on electricity systems, water supplies, and regional infrastructure.
Ulanqab Moves Into the AI Infrastructure Spotlight
Ulanqab is home to about 1.5 million people and sits about two hours west of Beijing by train. Since 2016, nearly 100 data centers have either opened there or begun construction.
The scale of planned investment is large. Chinese companies have pledged projects with a combined estimated capacity of 12.5 gigawatts in the city, according to a research note published by Goldman Sachs last week. More than 70 percent of those commitments were announced in just the last year.
That makes Ulanqab one of the fastest growing compute clusters in Asia. The comparison is striking: OpenAI's $500 billion Stargate Project is set to reach 10 gigawatts of total capacity when it is complete.
The companies involved also show how the market is changing. DeepSeek is reportedly building a massive AI data center in Ulanqab, and ByteDance, Alibaba, and Xiaohongshu are also building there. For years, Chinese AI companies spent less on physical infrastructure than American peers, even while releasing popular AI models with strong capabilities. Ulanqab suggests that pattern is shifting.
Why This Inner Mongolia City Works for AI
Several basic conditions make Ulanqab attractive for an AI data center. The city is located at high elevation on the Inner Mongolian Plateau, where long, cold winters reduce the energy required for cooling. That matters because data centers must manage heat continuously.
Its location also helps. Ulanqab is relatively close to Beijing, so data can move to populous regions with less delay than it would from more remote western hubs. Two dedicated fiber optics cables built in 2017 and 2019 reduced average latency speeds to less than five milliseconds, enough to support real-time data exchanges such as AI inference.
Cost may be the strongest pull. Electricity is cheaper in Inner Mongolia than almost anywhere else in China. That lower price is tied to both rapid growth in wind and solar energy and the region's abundant coal supply.
For AI companies, those conditions line up with two different needs:
- Training: A model training run can take months and does not require much real-time tinkering, making latency less of a problem.
- Inference: As paying users increase, companies need data centers close enough to serve applications without excessive delay.
That helps explain why a region once better suited to backup storage is now becoming relevant to front-line AI workloads.
From Backup Storage to AI Training
Inner Mongolia has been part of China's data center map for at least a decade. Huawei built its first data center in Ulanqab in 2016, and Apple followed three years later.
In 2021, the area was designated as one of the main hubs of a country-wide government project called “Eastern Data, Western Compute.” The plan aims to place data centers in China's western hinterlands.
At first, the distance from China's populous eastern coast created a problem. Higher latency made these facilities less useful for applications that needed fast interaction with users. As a result, the data centers were initially used largely for backup storage.
AI changed the value of that infrastructure. “With the rise of AI in 2022, there was the realization that actually, those remote data centers could be well-utilized for model training,” says Andrew Stokols, a professor at Singapore Management University who studies China's compute infrastructure.
Ulanqab also appears to differ from some other Chinese data center hubs. Stokols says its growth seems to have been driven more by commercial demand than by government-led investment. As Chinese AI startups including DeepSeek, Moonshot AI, and Zhiput AI attract more paying users at home, the business case for nearby inference infrastructure becomes clearer.
Renewable Energy Is Part of the Pitch
The Chinese government sees another advantage in Inner Mongolia: data centers could help use excess renewable energy. Damien Ma, director of Carnegie China, a Singapore-based research center, says there is a positive correlation between Chinese regions with the most unused renewable energy and those that have built the most data centers.
In that view, the strategy solves two problems at once. China can expand data center construction while also creating more demand for renewable power.
Envision, one of the largest Chinese manufacturers of wind turbines, announced this month that it will build a 2 gigawatt AI data center in Ulanqab connected directly to the company's own clean power supply. That project reflects the broader logic behind pairing compute growth with renewable generation.
But the energy story is not simple. Stokols found in his research that about 37 percent of electricity in Ulanqab still comes from coal. Data centers must operate around the clock, and operators have traditionally favored fossil fuels because of their reliability.
“Inner Mongolia has long been the West Virginia of China. It’s a coal country,” says Ma.
The region is racing to replace coal with wind and solar, but it remains unclear how quickly that transition will move or how complete it will be. For now, the data centers rising in China's remote highlands still depend on coal to some extent. Ma says that “in three years, maybe it will be completely powered by renewables.”
Water May Be the Hardest Constraint
The most immediate challenge may not be power, but water. Ulanqab is about as dry as Denver and receives roughly 14 inches of rain each year.
The local government is already struggling to provide enough water for residents before many planned data center projects are even operating. Last month, the local water company in Ulanqab had to turn off several waterworks for seven hours each night to reduce peak demand.
The data centers being built in the city will need less water in winter. Weather data from the local government of Ulanqab shows they require additional water for cooling during only two months out of the year. Even so, the amount of new infrastructure planned for the region could create a serious environmental challenge.
That is the central tension in Ulanqab's rise. The city has the power prices, climate, and connectivity that AI companies want. It also has limited water and an energy system still linked to coal. China's AI data center boom is not only a story about compute capacity; in Ulanqab, it is also a test of whether the physical systems beneath AI can scale without overwhelming the place that hosts them.