China's inaugural technical report on the framework for gigawatt-scale open AI computing centers has been officially released. This report was led by a prominent company based in Luoyang, which contributed its key technical expertise in two critical areas.
Avic Optoelectronics Technology Co., Ltd announced that the "Gigawatt (GW)-Scale Open AI Computing Center Framework Technical Report," which it spearheaded, was formally launched at the 2026 Open Compute Technology Summit. The company provided crucial technical outputs within the report's two core sections—high-speed interconnection and cooling systems—thereby contributing "Luoyang expertise" to the nation's computing power infrastructure development.
The Rise of GW-Scale AI Computing Hubs
As large model parameters continue to evolve and AI agents are deployed at scale, massive data exchange has become standard practice. Gigawatt-scale AI computing centers are now recognized as the foundational core for next-generation computing infrastructure. This technical report, grounded in global trends for GW-scale AI compute construction, establishes five underlying technological pillars: computing power, high-speed interconnection, storage, high-power power supply, and comprehensive liquid cooling. It serves as a systematic guidance document for the development of ultra-large AI computing centers nationwide.
Pioneering High-Speed Interconnection Solutions
In the high-speed interconnection domain, Avic Optoelectronics focused on the development trends for GW-scale centers. The report analyzes four interconnection scenarios: compute nodes, switch nodes, intra-cabinet, and inter-cabinet. By assessing transmission requirements to determine architectural evolution paths, it evaluates representative product solutions such as connectors, copper cables, and PCBs. Furthermore, it addresses the challenges of long and short-distance transmission within ultra-node whole cabinets, comparing the pros and cons of copper and optical interconnects. The report elucidates a "co-evolution of optical and copper, leveraging their complementary strengths" approach, pointing the way for the evolution of next-generation interconnection technologies.
Advancing Cooling System Technologies
Regarding cooling systems, Avic Optoelectronics compared three mainstream solutions: air cooling, cold plate liquid cooling, and immersion liquid cooling. Considering current industry implementation, the report identifies cold plate liquid cooling as the dominant application at this stage and provides a systematic breakdown of its technical architecture. In response to the increasing thermal density challenges posed by continuously growing AI models, it offers an in-depth comparison of the adaptation characteristics of cold plate versus immersion cooling. By forecasting development trajectories based on GW-scale construction needs, the report supplies theoretical and technical foundations for the large-scale deployment of thermal management solutions in super-large AI computing centers.
Establishing Industry Standards
This release follows the company's leadership in publishing the "AI Computing Center High-Speed Interconnection Technology White Paper (2026)" earlier this year. Once again targeting GW-scale AI computing center construction, Avic Optoelectronics has systematically organized the technology roadmap, product selection, and engineering implementation standards for optical and copper interconnects. This effort aims to establish a unified technical benchmark for the industry's high-speed interconnection sector.
As a leading enterprise in high-end interconnection within its region, Avic Optoelectronics plans to continue its deep involvement across the entire GW-scale open AI computing industry chain. The company intends to persistently iterate on interconnection innovation technologies, further participate in refining the national AI computing industry standard system, and leverage Luoyang's advanced manufacturing advantages to collaborate with industry chain partners on joint technology research and ecosystem development. These efforts are aimed at accelerating the high-quality growth of the open computing industry.
Comments