China’s state-backed telecom giants are speeding up the deployment of domestically developed AI chips in their data centers, in a nod to Beijing’s push for technological self-sufficiency and reduced reliance on foreign suppliers. China Mobile recently said it plans to build the country’s largest and most advanced artificial-intelligence computing network by 2028, a project that will run entirely on domestically produced chips.
China’s state-backed telecom giants are speeding up the deployment of domestically developed artificial intelligence chips in their data centers, in a nod to Beijing’s push for technological self-sufficiency and reduced reliance on foreign suppliers.China Mobile recently said it plans to build the country’s largest and most advanced AI computing network by 2028, a project that will run entirely on domestically produced chips.
The company pledged to double its overall investment in AI, saying it will build a GPU cluster of 100,000 accelerator cards with total computing power exceeding 100 exaFLOPS nationwide, creating a key platform to support China’s AI development.
By the end of last year, the company’s self-built intelligent computing capacity — excluding leased resources — had reached 29.2 exaFLOPS at FP16 precision, according to its 2024 annual results.
The company plans to invest 37.3 billion yuan ($5.1 billion) in computing power this year, Chief Executive He Biao said at the company’s 2024 annual results briefing on March 20. This will bring its accumulated general-purpose computing capacity to 8.9 exaFLOPS and AI computing power to more than 34 exaFLOPS at FP16 precision.
Although the company didn’t disclose which AI-chip vendors it plans to work with, its ambitious plan presents opportunities for companies such as Huawei, Cambricon Technologies, Biren Technology, Moore Threads and Tianshu Zhixin.
In fact, some of these companies already have partnerships with the telecom giant.
China Mobile has adopted general-purpose GPU products supplied by Biren Technology at its intelligent computing center in Hohhot, Inner Mongolia. Cambricon Technologies took part in building the world’s largest single-cluster intelligent computing center operated by its intelligent computing center in Harbin.
According to China’s national procurement portal, China Mobile’s 2025–26 tender for AI computing devices included a CANN ecosystem package tied to Huawei’s Ascend chips. The winners — Henan Kunlun Technology, Sichuan Huakun Zhenyu and Wuhan Yangtze Computing Technology— are Huawei partners supplying servers powered by Ascend processors.
China Unicom and China Telecom are making similar moves.
Last month, state broadcaster CCTV reported that China Telecom built a massive data center in Xining, the capital of the western province of Qinghai, supplied mainly by Alibaba’s chip unit T-Head and other Chinese companies including MetaX, Biren Technology, Moore Threads and Enflame.
A notable example of China Telecom’s efforts is its collaboration with Huawei in Shaoguan, Guangdong Province, where the two companies unveiled a commercial “Ascend Supernode” computing center powered by Huawei’s Ascend AI architecture.
These efforts have gained momentum as Chinese authorities call for building “self-reliant and domestically controlled” intelligent computing infrastructure amid escalating trade and geopolitical tensions with the US (see here).
— Setbacks for foreign suppliers —
In the past, state-backed telecom giants occasionally purchased chips from foreign suppliers to meet customer demand.
For leading foreign suppliers such as Nvidia, its H20 chips were favored more by major private technology companies including ByteDance, Alibaba and Tencent ahead of US export restrictions in April.
However, Nvidia’s sales in China suffered a setback after the country’s internet authority summoned the company, citing concerns over potential vulnerabilities, including “loopholes and backdoor” risks in the chips (see here). Domestic regulators reportedly urged private companies to avoid using the H20 for sensitive or government-related projects.
Nvidia Chief Executive Jensen Huang recently lamented a sharp decline in the company’s AI chip market share in China (see here).
The push for self-sufficiency is adding pressure on other foreign chipmakers as well, especially those that have already come under government scrutiny.
In a recent statement, US memory-chip giant Micron said its data-center business in China was affected by a 2023 cybersecurity review conducted by the country’s internet regulator (see here).
Reuters reported that the chipmaker plans to exit the server-chip business in China but will continue supplying semiconductors for the automotive and mobile sectors.
- Analysis by Olivia Wang.
Please email editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.