China’s push to shift AI computing inland is becoming an economic and strategic play, with Ulaanqab in Inner Mongolia emerging as one of the country’s most important low-cost data center hubs and potentially saving operators as much as 7 billion yuan a year on electricity for a single large project.
Ulaanqab becomes China AI data center hub
The scale of the advantage helps explain why the city is drawing major technology investment. For AI infrastructure, power is now one of the biggest operating costs, and Ulaanqab combines cheap electricity, cold weather and direct fiber links to Beijing in a way that makes it unusually attractive for compute-heavy workloads. That matters for China’s broader effort to build out AI capacity while keeping costs down and reducing strain on coastal power grids.
Sentiment Indicators
Envision Group’s Envision Galaxy Campus illustrates the economics. The project is expected to exceed 2 gigawatts of power capacity and will include 120,000 square meters of facilities able to house as many as 1 million AI computing cards. At full load, its annual electricity use could reach 17.5 billion kilowatt-hours. With local power priced around 0.30 yuan per kWh versus 0.60 yuan to 0.70 yuan in eastern coastal provinces, the implied annual saving ranges from 5.25 billion yuan to 7 billion yuan.
Those savings are not just a line-item benefit. They are material to the economics of training and running large AI models, where electricity, cooling and network latency can decide whether a site is commercially viable. Ulaanqab’s average annual temperature of 4.3 degrees Celsius allows outside air cooling for as much as 10 months a year, trimming the need for energy-intensive mechanical cooling. The city’s 4.2-millisecond latency to Beijing also lets companies keep research and business functions close to the capital while pushing heavy compute to a lower-cost location.
The move fits Beijing’s “East Data, West Computing” policy, which is designed to divert data processing away from crowded coastal regions toward inland areas with cheaper power and more land. That policy is as much about industrial efficiency as it is about technology leadership: China wants to expand AI capacity without locking itself into the high power costs and infrastructure bottlenecks that have become a constraint for U.S. cloud and AI operators as well.
Investors should see the Ulaanqab build-out as part of a wider China AI infrastructure trade. The beneficiaries include power producers, renewable energy developers, fiber and networking suppliers and AI hardware vendors such as Nvidia, Advanced Micro Devices and Taiwan Semiconductor Manufacturing, all of which stand to gain from continued build-out of compute capacity. The losers are coastal data center operators facing weaker cost competitiveness, and potentially overseas rivals if China’s lower-cost infrastructure helps accelerate domestic AI deployment despite export controls and geopolitical friction.
The bull case is that Ulaanqab becomes a durable national node for AI compute, lowering the marginal cost of inference and training and supporting faster scaling by Chinese platforms including Huawei, Alibaba and ByteDance. The bear case is that the economics depend on continued access to chips, power and network capacity, while policy or trade restrictions could still slow deployment. For now, though, the message is clear: in China’s AI race, cheap electricity and cold air are becoming as important as silicon.
| Entity | Gains | Losses |
|---|---|---|
| Ulaanqab data centers | ▲Lower operating costs | ▼Higher build-out dependence |
| Envision Group | ▲Cheaper AI compute | ▼Capital-intensive expansion |
| Chinese AI firms | ▲Faster scaling | ▼Coastal cost pressure |
| Coastal data center operators | ▲None | ▼Loss of price competitiveness |
