电子行业2025年中期投资策略：AI投资的新范式

Group 1: Core Insights - The report emphasizes the long-term growth potential of computing power, exploring the latest developments in GPUs and ASICs, and identifying changes in domestic computing power and AI terminals [4][5][7] - Recent performance of AI hardware and software stocks in the US market has reached new highs, driven by Nvidia's better-than-expected earnings and the increasing demand for AI-enabled internet applications [4][20][24] - The computing power upgrade is driven by two main routes: speed and power, with advancements in PCB upgrades, server architecture changes, and the necessity of liquid cooling for increased chip power consumption [4][5][7] Group 2: Overseas Computing Power - The report highlights the rapid growth in inference demand, which is expected to create a return on investment (ROI) loop for AI investments, with Nvidia's product iterations accelerating in response to this demand [4][20][24] - Nvidia's recent earnings report showed a revenue of $44.1 billion for FY25Q3, a year-on-year increase of 69%, indicating strong market demand for AI computing power [24] - The global ASIC market is projected to grow from $6.5 billion in 2024 to $15.2 billion by 2033, with a compound annual growth rate (CAGR) of 12.8%, reflecting the increasing importance of ASICs in the computing power landscape [29][75] Group 3: Domestic Computing Power - Domestic AI models like Doubao and DeepSeek are accelerating the development of Chinese large models, with significant updates from various domestic companies since 2025 [5] - The report notes that domestic cloud computing firms are increasing their investments in computing power infrastructure, although short-term supply may not meet the rapidly growing demand [5] - The report identifies key domestic companies in the computing power ecosystem, including chip manufacturers like SMIC and Cambrian, which are making significant strides in adapting to the domestic computing power landscape [5][75] Group 4: AI Terminals - The report discusses the ongoing structural innovations in AI terminal hardware, such as smartphones and smart glasses, with a particular focus on the rising market for AI/AR glasses [7] - The interaction modes and functionalities of AI glasses are currently limited, but the integration of AR features is expected to enhance user experience significantly [7] - The report expresses optimism about the long-term narrative of the AI industry, highlighting the strong performance of Nvidia and the rise of domestic computing power breakthroughs as key investment opportunities [7] Group 5: Investment Recommendations - The report suggests focusing on specific sectors within the computing power chain, including servers, PCBs, CPOs, copper cables, and power supply systems, where domestic companies have established advantages [8] - Key companies to watch include Industrial Fulian and Huajin Technology in the server space, and Corechip and Cambrian in the computing chip sector [8] - The report also highlights the importance of supply chain partners in the ASIC market, indicating a shift towards a more competitive landscape with multiple players emerging [8][75]