Via klook.com
Envision Group's Galaxy Campus in Inner Mongolia can house 1 million AI accelerators, powered by renewable energy in a region where the average temperature barely cracks 40°F.
A pastoral city in Inner Mongolia that most people outside China have never heard of just became one of the most important coordinates on the global AI map. Ulanqab is now home to China’s largest artificial intelligence industrial park.
The centerpiece is Envision Group’s Galaxy Campus, which became operational around August 6. The facility’s main building spans 120,000 square meters and is designed to house up to 1 million AI accelerators, with a target compute capacity exceeding 1 million petaflops.
Why a city in the middle of nowhere #
Ulanqab’s appeal is deceptively simple: it’s cold, it’s cheap, and it’s windy. The city’s average temperature hovers around 4°C (roughly 39°F), which dramatically reduces the cooling costs that typically eat into data center margins.
The region also sits on abundant wind and solar resources, which feeds directly into Envision’s broader pitch. The entire Galaxy Campus aims to surpass 2 gigawatts of energy capacity, sourced primarily through integrated renewables.
This setup aligns with Beijing’s “East Data, West Computing” national strategy, which routes data-intensive workloads from expensive eastern cities to resource-rich western regions. Ulanqab has attracted 89 data center projects to date, combining for over 5 million contracted server racks. Alibaba, Huawei, and Tencent all already operate infrastructure in the region.
Envision’s global ambitions and local competition #
Envision Group has laid out plans to build 5 GW of green AI computing capacity in desert and arid regions globally by 2030. The Ulanqab facility is the flagship.
DeepSeek, the Chinese AI startup that grabbed international attention earlier this year with its efficient open-source models, is reportedly in discussions for approximately 1 GW of AI data center capacity in Ulanqab. The company currently leases infrastructure from existing providers in the area.
What this means for the global AI race #
As US export controls continue to restrict the flow of advanced chips to Chinese firms, Beijing’s response has been to build massive infrastructure that maximizes the output of whatever hardware is available. A facility designed for 1 million accelerators doesn’t need each individual chip to be best-in-class if the sheer volume of compute compensates.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our