Wuwen Xinqiong and Huahuan Electronics Sign Strategic Cooperation on Domestic Heterogeneous-Compute AI Infrastructure

Guo Meng, Party secretary and chairman of Huahuan Electronics, and Zeng Shulin, Shanghai general manager of Wuwen Xinqiong, signed the agreement. Zhen Jia, Party secretary and chairman of Qingkong Group; Wang Yu, a tenured professor in Tsinghua University's Department of Electronic Engineering and an IEEE Fellow; and Li Boxun, CTO of Wuwen Xinqiong, witnessed the signing.

The agreement comes as large models and agent applications push AI infrastructure from a single compute-resource focus toward coordination among computing, networking, software platforms and model services. Huahuan Electronics is a national-level specialized and sophisticated "little giant" enterprise with historical ties to Tsinghua University's Department of Electronic Engineering and the Tsinghua technology industry system. It has worked in information communication and optical networking for more than 30 years, and its optical transmission equipment has long served China Mobile, China Telecom and China Unicom, as well as dedicated communication networks in power and transportation. Wuwen Xinqiong also originated from Tsinghua's electronics department and describes itself as an AI-native infrastructure provider. Its technologies include multi-heterogeneous computing, software-hardware collaboration and autonomous AI, and it has built the Agentic Infra platform and Agentic MaaS large-model service platform, serving more than 1,000 AI companies and research institutions.

Under the agreement, Wuwen Xinqiong will contribute heterogeneous-computing management and AI software platform capabilities, while Huahuan Electronics will provide network communications, hardware R&D, engineering implementation and operations and maintenance. The two companies plan to explore infrastructure collaboration for domestic heterogeneous computing, aiming for end-to-end complementary capabilities from compute scheduling to network transmission and from software platforms to hardware engineering.

For intelligent computing center solutions, they intend to jointly advance solution design, software and hardware adaptation, integrated delivery, and operations and maintenance. Wuwen Xinqiong has built a cross-domain training system for heterogeneous clusters that uses compute-communication overlapping pipeline scheduling, adaptive communication flow scheduling and heterogeneous compute scheduling. QbitAI reported that the system has been deployed nationwide to reach more than 37,000P of computing power and cover 16 mainstream chips. Huahuan Electronics will provide high-speed interconnection and stable transmission support in optical transmission and network access. The partners also plan to establish a regular cooperation mechanism for technical validation and commercialization.

The companies will further explore an infrastructure supply model that uses tokens as standardized output and is driven by AI application demand, focusing on an optimization loop of domestic large models plus domestic chips. Wuwen Xinqiong's Agentic MaaS platform is intended for large-scale, high-quality token production, and the company says it has reduced inference costs tenfold over the past year. Huahuan Electronics' hardware R&D, network communications and engineering operations capabilities are expected to provide engineering support for building and operating a token factory.

Both sides said they hope to follow a path from domestic chips to heterogeneous-computing infrastructure to token services to enterprise AI applications, creating a replicable model for building and operating intelligent computing centers. They aim to turn domestic software and hardware capabilities into AI productivity for real business use, and to support large-scale AI adoption with more efficient, stable and open infrastructure.