英伟达新映射：大震荡来袭，国产算力自主可控升级大势不改

Core Viewpoint - The long-term logic of AI computing power remains unchanged, with accelerated domestic substitution in progress, driven by the latest performance of Nvidia, which has significant implications for global capital markets [1][2]. Group 1: Market Performance and Trends - Nvidia's revenue and guidance remain strong, with its stock price initially rising over 5% on the earnings release day but ultimately closing down 3.15%, reflecting market concerns about sustainable future growth amid external uncertainties [2]. - The decline in the US tech sector has impacted the A-share market, with the computing power concept sector dropping 3.38% on November 21, resulting in a net outflow of 14.12 billion yuan [3]. - The AI chip sector also fell by 4.33%, with a net outflow of 1.685 billion yuan, indicating a reevaluation of valuations and fundamentals in the computing power sector [4]. Group 2: Domestic Market Dynamics - By Q3 2025, the communication industry is expected to experience structural growth driven by strong AI computing power demand, with overseas markets as a core engine [5]. - Domestic internet companies are showing varied capital expenditure trends due to chip supply constraints, with Alibaba planning to invest over 380 billion yuan in AI and cloud infrastructure over three years, while Tencent's capital expenditure decreased by 24% year-on-year [5]. - The proportion of domestic chip suppliers in China's AI server market is projected to rise from 37% in 2024 to 40% by 2025, indicating substantial progress in domestic substitution [6]. Group 3: Technological Advancements - Domestic chip manufacturers have transitioned from conceptual phases to performance realization, with significant revenue growth reported by companies like Cambricon and Haiguang Information [6]. - The domestic computing power ecosystem is forming a more complete division of labor, with advancements across all segments from chip design to data center deployment [7]. - The evolution of technology is shifting from hardware accumulation to efficiency enhancement, with research indicating that the maximum capability density of large models is expected to double approximately every 3.5 months [8]. Group 4: Industry Collaboration and Policy Support - The collaboration across the computing power industry chain is strengthening, supported by national policies and funding for research projects, creating a favorable environment for domestic chip development [7]. - System-level innovations, such as the adoption of supernode technology, are helping domestic companies optimize architectures to meet the computing power demands of large model training [9]. Group 5: Investment Opportunities - AI computing power has become a major growth engine in the communication industry, with investment opportunities emerging in various segments, particularly in the optical module sector [10]. - The demand for storage technology is rapidly evolving due to AI training requirements, pushing domestic storage chip manufacturers to accelerate their advancements [11]. - The investment focus in the computing power industry is shifting from infrastructure construction to application innovation, reflecting the maturation of the sector [11].