Workflow
AI算力网络基础设施建设
icon
Search documents
从WAIC2025看行业发展趋势:昇腾384首亮相,国产算力催化加速
Investment Rating - The industry investment rating is "Outperform the Market," indicating that the industry index is expected to perform better than the benchmark index over the next 6-12 months [11]. Core Insights - The report highlights the launch of Huawei's Ascend 384 super node, which represents a shift in the domestic AI industry from a chip-centric approach to a system-coordinated optimization strategy. This development showcases the robust growth momentum of the domestic computing power industry chain [3][6]. - The Ascend 384 super node, consisting of 12 computing cabinets and 4 bus cabinets, integrates 384 Ascend 910C NPUs and 192 Kunpeng CPUs, achieving a theoretical performance close to twice that of NVIDIA's GB200 NVL72. It provides 300 PFLOPS of BF16 computing power, addressing the bottlenecks in large model training through high-speed interconnect technology [3][6]. - The report emphasizes the strong order fulfillment capability of the Ascend platform, which has been widely applied in high-computing scenarios across various sectors, including finance, power, transportation, and government [3][6]. Summary by Sections Industry Development Trends - The report notes that the domestic AI computing power industry is expanding, driven by the increasing demand for controllable AI infrastructure amid high-end chip shortages and export restrictions. The success of the Ascend 384 super node demonstrates the feasibility of achieving a closed-loop in the domestic computing power industry chain [3][6]. Investment Recommendations - The report suggests prioritizing investments in the domestic AI computing power network infrastructure, focusing on telecom operators such as China Mobile, China Telecom, and China Unicom, as well as server and switch equipment manufacturers like ZTE, Unisoc, Inspur, and Ruijie Networks [3][6]. Market Dynamics - The report indicates that significant investments from major players like Alibaba, which plans to invest over 380 billion yuan in cloud computing and AI hardware infrastructure over the next three years, will open up larger market opportunities for domestic computing power solutions [3][6].