China Unicom Opens AI Data Centre in China with Local Chips
Chinese telecom China Unicom confirms the launch of a new data centre in Xining, Qinghai province.
Announced on state broadcaster CCTV and reported by Reuters, the facility is described as "massive" and is built on around US$390m of investment. The project is presented as a showcase for China’s push to deploy AI and high-performance computing using domestically developed technology.
The data centre is designed to run on local processors instead of foreign-made chips, aligning with Beijing’s policy to reduce reliance on imported hardware.
It currently operates with 23,000 AI chips made by Chinese firms and provides 3,579 petaflops of computing power. Once completed, the centre is expected to deliver 20,000 petaflops. A petaflop measures a computer’s processing speed, with one petaflop equal to one quadrillion floating-point operations per second.
Domestic chips at the core
The majority of processors deployed in the centre come from Alibaba’s semiconductor unit T-Head, which supplies about 72% of the total. Other contributors include MetaX, Biren Tech and Zhonghao Xinying. Future upgrades will add chips from Tecorigin, Moore Threads and Enflame, further expanding the domestic supplier base.
China Unicom, Alibaba, Biren, MetaX, Enflame, Tecorigin and Zhonghao Xinying did not respond to Reuters’ requests for comment, nor does the announcement include details on the site’s scale in terms of power capacity or floor space.
It has not been specified whether the compute performance figures are calculated in FP64, the double-precision floating-point standard used in high performance computing and the Top500 supercomputer list, or FP8, which is often used for AI training and inferencing workloads.
Local policy pushes domestic adoption
China is directing operators of public data centres to increase reliance on locally developed hardware.
In August, the Chinese government issued a mandate requiring publicly owned facilities to source more than 50% of their chips from domestic suppliers. The strategy supports the country’s efforts to strengthen its semiconductor ecosystem and reduce exposure to export restrictions.
The move gains further weight as China’s internet regulator has reportedly imposed a ban on the country’s largest technology companies, preventing them from purchasing Nvidia’s AI processors. Nvidia chips are widely used in global AI systems but have become the subject of trade restrictions between Washington and Beijing.
The new data centre therefore represents more than just a facility launch. It signals a broader policy direction that ties infrastructure expansion with the growth of the local semiconductor sector.
Alibaba steps up chip development
Alongside its contribution to the Xining facility, Alibaba is advancing its own semiconductor programme. The company revealed earlier this month that it is developing a new AI inferencing chip.
Testing is already under way and the processor has been manufactured domestically. This development fits into the government’s wider plan to strengthen capabilities in AI-specific chip design and fabrication.
While details about deployment timelines remain limited, the addition of new AI chips signals Alibaba’s increasing role in supplying domestic compute for data centres. As the leading chip provider for the China Unicom facility, its T-Head unit underlines the link between major cloud operators, telecoms firms and the emerging domestic chip ecosystem.
Building for AI and HPC demand
China Unicom’s new site illustrates the scale of investment directed toward AI and high-performance computing capacity in China.
With computing power expected to scale up to 20,000 petaflops, the centre contributes to the country’s strategy of expanding its infrastructure footprint while reducing reliance on foreign chip imports.
By prioritising domestically made processors from multiple local developers, the project aligns with policy goals and creates a testing ground for China’s growing chip sector.
