China speeds AI data-center construction with stackable prefabricated modules

China is using stackable prefabricated modules to accelerate data-center construction, NBC News reported November 13 from Ulanqab, Inner Mongolia. The city’s rapidly changing landscape offers a ground-level view of the country’s AI infrastructure push: CBS News reported a day earlier that dozens of facilities are under construction across the region, supported by up to 30,000 workers brought from elsewhere in China.

Building with modules and a large workforce

Prefabrication is one practical contributor to the pace. NBC describes modules that can be stacked together, helping speed construction as China adds computing capacity to compete with the United States. Its reporting places that activity in a wider geographic shift toward less-populated regions with large supplies of renewable energy, rather than simply near the country’s biggest population centers.

On the ground, the work extends well beyond isolated server buildings. CBS journalists drove miles along new roads with multistory data centers rising on both sides. From the road, they could see racks intended to hold servers and chips, along with cooling and exhaust equipment. Trucks, cranes and thousands of workers filled the construction area.

Cheap electricity is not all renewable

Ulanqab combines several advantages for developers. Electricity is abundant and inexpensive, while the cold climate helps reduce cooling costs. Land and labor are also cheaper in the remote region. Wind turbines and solar installations contribute to its energy supply, but fossil fuels remain the main source of power, according to CBS. The buildout is not running on renewables alone.

CBS also relays an estimate from experts cited by the Financial Times that Chinese data centers cost about one-fifth as much to build as their U.S. counterparts, partly because of land and labor costs. That is an attributed construction-cost comparison—not a measurement of computing performance or the cost of running a particular AI model.

The companies active in Inner Mongolia include Huawei, Alibaba, ByteDance, DeepSeek, Baidu, Z.ai, ZData and VNET. Some facilities have already moved beyond construction: CBS describes completed sites with fences, landscaped grounds and avenues named after AI companies, alongside the buildings still going up. The regional picture therefore includes both operating facilities and unfinished projects.

Local benefits, with limits on dissent

Residents interviewed by CBS welcomed parts of the transformation. Baletu, a 72-year-old farmer who has grown potatoes and corn locally for more than 40 years, said the development had not affected his farmland. A businesswoman selling lunch outside the construction site emphasized job opportunities for young people. Those comments describe individual experiences, not a comprehensive measure of local support.

The political setting also complicates comparisons with American community opposition. CBS notes that China’s government generally does not tolerate protests and tightly controls media messaging. Its account of fewer hurdles for Chinese developers sits beside that constraint; an absence of visible resistance should not be read as equivalent to freely expressed public agreement.

Advanced-chip access remains a separate constraint: CBS identifies restrictions on high-end chip sales as a drag on Chinese AI development. Henry Wang, founder and president of the Center for China and Globalization, argues that electricity could ultimately become the decisive bottleneck—and that China currently has an advantage. That is a forecast, not a demonstrated outcome of Ulanqab’s construction boom.