Huawei partners win big as China Telecom invests US$1.7 billion in 40,000 servers
China Telecom has named the winners of a multibillion-dollar high-performance server procurement programme, handing a major victory to domestic CPU chip companies like Huawei Technologies as state firms rapidly replace US technology with domestic hardware. The deal, which covers China Telecom’s need
By Howard Liu
China Telecom has named the winners of a multibillion-dollar high-performance server procurement programme, handing a major victory to domestic CPU chip companies like Huawei Technologies as state firms rapidly replace US technology with domestic hardware.
The deal, which covers China Telecom’s needs for 2026 and 2027, called for 40,000 high-performance servers split into two packages, according to notices on its procurement platform on Tuesday. While the final price tag was not made public, tender documents put the budget ceiling at 11.55 billion yuan (US$1.7 billion).
Huawei’s expanding footprint was evident in the larger of the two packages, which covers 28,000 Arm-based servers tied to the company’s Kunpeng ecosystem. Unlike traditional x86 server chips dominated by US giants Intel and Advanced Micro Devices, Arm-based servers use a simplified chip architecture widely used in mobile devices.
While Huawei did not bid on the contract itself, the six winning companies were all publicly linked to the Kunpeng ecosystem, although the notices did not disclose the specific processors used in their bids. Such arrangements permitted Huawei to quietly capture major chunks of the government procurement market without its name showing up on the official supplier list.
The other part of the deal covered 12,000 C86 servers, a category associated with domestic computing platforms that are compatible with traditional x86 systems. The list of winners is a who’s who of Chinese tech giants, including ZTE, H3C, Inspur and Lenovo.
China Telecom’s order mirrors recent buying sprees by other big state-owned carriers.
In April, China Mobile launched a bid for nearly 63,000 servers, including over 40,000 running on Arm architecture, according to local telecoms outlet C114. Last year, China Unicom put out a call for 87,000 servers, with the vast majority built on ecosystems designed by Huawei and Beijing-based Hygon Information Technology.
As some of the country’s biggest buyers of data centre and cloud infrastructure, China Telecom, China Mobile and China Unicom hold substantial sway over the industry. Their shopping lists can help determine which of the domestic tech ecosystems get the cash, volume and real-world validation they need to grow their businesses.
The recent string of purchases by the industry’s big players has been driven by Beijing’s technology localisation campaign, known as Xinchuang, meaning information technology application innovation. What started out as a policy to replace foreign software and central processing units in standard government office computers has been broadening rapidly.
Spurred on by tight US export controls, China is overhauling its digital backbone, replacing foreign tech in databases, cloud infrastructure and artificial intelligence chips across critical sectors including banking and telecoms.
China’s Xinchuang hardware market is expected to reach 788.95 billion yuan in 2026, as replacement demand accelerates across eight key industries, Guotai Haitong Securities said in an April report, citing CCID Consulting.
Analysts have said that China’s domestic computing push was moving beyond policy-driven procurement and becoming more closely tied to real infrastructure demand.
“AI-related Xinchuang is no longer just an extension of traditional IT localisation, but a contest over control of the underlying infrastructure in national-level technology competition,” Soochow Securities said in a May report.
The same report said China’s AI infrastructure was shifting from simple GPU build-outs towards heterogeneous systems combining GPUs and CPUs, as agent-based AI workloads and reinforcement learning are increasing the demand for scheduling, input-output management and general-purpose computing.