证券研究报告 | 行业深度报告 2025 年 08 月 11 日 国内设备/算力/代工等板块业绩增长向好，关注存储/模拟等复苏态势 | | | 占比% | | --- | --- | --- | | 股票家数（只） | 507 | 9.9 | | 总市值（十亿元） | 10138.5 | 10.6 | | 流通市值（十亿元） | 8673.4 | 10.0 | 行业指数 半导体行业深度跟踪 TMT 及中小盘/电子 海外 CSP 云厂商纷纷上修资本支出，台积电上修 2025 年收入增速指引，整体 算力景气度有望延续，国内海光等亦持续释放业绩；海外半导体设备公司 25Q2 表现符合预期，国内设备公司签单和业绩表现向好；海外存储原厂受益于 HBM 等需求，国内存储模组和利基存储芯片公司将继续受益于涨价和库存改善等， 整体收入和盈利能力预计均保持边际复苏态势；海外 TI 表示大部分终端市场复 苏，国内模拟公司 25Q2 营收多数同环比改善。建议关注自主可控加速叠加业 绩向好的设备/算力/代工/等板块、景气周期边际复苏的存储/模拟等板块，同时 建议关注各科创指数和半导体指数核心成分股。 2、库存端：手机链 DOI 环比 ...

Core Viewpoint - The semiconductor industry is experiencing a recovery in demand, particularly in consumer electronics and AI applications, while inventory levels are improving and capital expenditures are being adjusted upwards by key players like TSMC and major overseas semiconductor manufacturers [1][2][4]. Demand Side - The consumer electronics sector is witnessing a recovery, with AI and automotive applications driving innovation. Global smartphone shipments in Q2 saw a year-on-year growth slowdown to 1%, while domestic shipments declined by 4%. PC shipments increased by 6.5% year-on-year in Q2, but growth is expected to weaken in H2 2025. Wearable devices, particularly AI glasses, are experiencing significant growth, with a 216% year-on-year increase in Q1 shipments [2][3]. - The automotive market in H1 2025 saw a 13.8% year-on-year increase in sales, with new models like Xiaomi's YU7 achieving substantial pre-order numbers [2]. Inventory Side - The Days of Inventory (DOI) for the smartphone supply chain has slightly increased, with terminal customer inventories remaining low. Major chip manufacturers like Intel reported a decrease in inventory levels, indicating a potential for increased shipments as demand rises [3]. Supply Side - TSMC is ramping up its advanced process production lines in the U.S., driven by strong demand from AI data centers. The capacity utilization rates for various foundries, including SMIC and UMC, have shown improvements, with SMIC reaching 92.5% in Q2 2025 [4][12]. - Major memory manufacturers are focusing capital expenditures on high-end memory products like HBM, with companies like Micron and SK Hynix increasing their capital spending forecasts for 2025 [4]. Price Side - Memory prices are showing signs of steady recovery, particularly for DDR4 and NAND Flash products. The analog chip sector is also expected to see potential price increases, which could positively impact the industry [6]. Sales Side - Global semiconductor sales reached $59.9 billion in June 2025, marking a 19.6% year-on-year increase. The Asia-Pacific region, excluding China and Japan, saw a 34.2% year-on-year sales growth, while sales in Japan declined by 2.9% [6]. Industry Chain Tracking - The semiconductor industry is experiencing marginal improvements in various segments, with a focus on companies benefiting from domestic control and recovery in demand. The design and IDM sectors are seeing growth driven by consumer demand and AI applications [6][17]. - The MCU market is recovering, particularly in consumer electronics and automotive sectors, with significant sales increases noted in Q2 2025 [7]. - The storage sector is seeing improved profitability among overseas manufacturers, with domestic companies also expected to recover in H2 2025 [8]. - The analog chip market is experiencing good demand trends, with many domestic companies reporting improved revenue in Q2 2025 [9]. Advanced Manufacturing and Testing - The demand for advanced processes remains strong, with TSMC maintaining a robust growth forecast for AI-related chips. The utilization rates for mature processes are also recovering, although customer inventory demand for Q4 2025 remains uncertain [12][13]. - Major international packaging and testing companies are forecasting significant revenue growth for Q3 2025, with domestic companies increasing investments in advanced packaging capabilities [13]. Equipment, Materials, and Components - Domestic semiconductor equipment manufacturers are showing positive trends in orders and revenue growth, while international firms express caution regarding future guidance due to geopolitical uncertainties [14]. - The push for domestic supply chain independence is accelerating, with ongoing efforts to reduce reliance on foreign components [14]. EDA/IP Sector - Chip design companies like Chipone are reporting record-high order backlogs, indicating strong demand in the EDA sector [15].

Core Viewpoint - The strong performance of leading optical module companies is driving a bullish trend in the optical module market, supported by significant growth in AI-related capital expenditures and technological advancements in AI infrastructure [1][3]. Group 1: Market Performance - The optical module market is experiencing a robust rally, influenced by the impressive earnings reports from leading companies in the sector [1]. - The Communication ETF (515880) and the Growth Enterprise Market AI ETF (159388) have both seen increases of over 2.5% [2]. Group 2: Industry Growth - The global market for optical modules is projected to grow at a compound annual growth rate (CAGR) of 22% from 2024 to 2029, potentially exceeding $37 billion by 2029 [3]. - Domestic and international cloud service providers are significantly increasing capital expenditures, leading to rapid expansion in AI infrastructure demand, particularly for high-speed optical modules [3]. Group 3: Company Performance - Industrial Fulian reported record high revenue and net profit in its 2025 semi-annual report, benefiting from growth in AI business [1]. - Several optical module companies have disclosed strong performance in their semi-annual earnings forecasts, contributing to a high level of industry optimism [4]. Group 4: Technological Advancements - NVIDIA's H20 sales recovery and the rapid ramp-up of the Blackwell GPU are expected to enhance the demand for optical modules as major AI companies accelerate their infrastructure development [5]. - The Blackwell GPU is noted for its fastest ramp-up speed in NVIDIA's history, further driving the demand for optical modules [5]. Group 5: ETF Composition - The Communication ETF (515880) has over 40% exposure to optical modules, positioning it to benefit from the ongoing development of AI computing capabilities [6]. - The top ten constituents of the Communication ETF include key players in the optical module sector, such as Zhongji Xuchuang and Xinyi Sheng, which have significant weightings [8]. - The Growth Enterprise Market AI ETF (159388) tracks an index focused on the AI industry chain, providing greater elasticity due to its structure [9].

Investment Rating - The report maintains a "Buy" rating for the company [2] Core Views - The company's revenue for the first half of 2025 reached 5.46 billion yuan, a year-on-year increase of 45%, driven by the rising demand for domestic high-end chips [6] - The company is planning to merge with Zhongke Shuguang to enhance its data center infrastructure product offerings [6] - The domestic demand for local computing chips is expected to increase due to scrutiny on NVIDIA's H20 chips in China, positioning the company as a preferred supplier [6] - Revenue projections for 2025, 2026, and 2027 are estimated at 12.9 billion, 16.95 billion, and 22.1 billion yuan respectively, with corresponding net profits of 2.934 billion, 3.97 billion, and 5.388 billion yuan [7] Financial Performance - In Q2 2025, the company reported a revenue of 3.06 billion yuan, a year-on-year increase of 41% and a quarter-on-quarter increase of 28% [6] - The gross margin for Q2 2025 was 59.3%, down 4.5 percentage points year-on-year [6] - The company's inventory reached a historical high of 6 billion yuan by the end of Q2 2025 [6] Market Position - The company is expanding its partnerships with well-known server manufacturers such as Lenovo and Xinhua San [6] - The merger with Zhongke Shuguang aims to strengthen the supply chain and enhance product offerings [6] Future Outlook - The company is expected to see a gradual increase in market share for its CPU products and upgrades in its DCU products to capture the high-end computing market [6] - The projected earnings per share (EPS) for 2025, 2026, and 2027 are 1.26, 1.71, and 2.32 yuan respectively [7]

Investment Rating - The industry investment rating is "Overweight" (maintained) [1] Core Viewpoints - The recent security risks associated with Nvidia's H20 chip have drawn regulatory attention, highlighting the importance of a self-controlled supply chain in China [4][5] - The incident reflects the ongoing technological competition between China and the US, which may temporarily suppress capital expenditure from domestic internet companies but could ultimately drive the development of a stable and healthy AI industry in China [5] - Domestic internet companies are expected to accelerate their transition to local supply chains due to the uncertainties surrounding Nvidia's H20 supply [6] Summary by Sections Industry Trends - The AI chip industry in China is rapidly developing, with products like Huawei's Ascend 910B/910C surpassing Nvidia's H20 in computing performance [7] - Domestic manufacturers are making significant progress in technology and ecosystem development, with companies like Moore Threads and Huawei achieving breakthroughs [7] Supply Chain Challenges - The supply bottleneck in China's AI industry is becoming more pronounced, with domestic chip production facing limitations due to EUV export restrictions and TSMC's foundry constraints [7] - The production yield of domestic semiconductor equipment remains low due to developmental bottlenecks and lack of experience [7] Investment Recommendations - Beneficiaries in the AI chip sector include companies like Cambricon [8] - In the wafer foundry segment, companies such as SMIC and Hua Hong Semiconductor are recommended [8] - For lithography equipment, companies like Fuchuang Precision and Maolai Optics are highlighted [8] - Other recommended companies in front-end process equipment include North China Innovation and Zhongwei Technology [8] - In advanced packaging, companies like Chipbond and Huahai Qingke are suggested [8] - EDA beneficiaries include companies like Huada Jiutian and Gai Lun Electronics [8]

Core Viewpoint - The article discusses the illegal export of Nvidia's advanced AI chips, particularly the B200 GPU, to China despite U.S. export restrictions, highlighting the emergence of a black market for these products [1][2][3]. Group 1: Nvidia's AI Chips and Black Market Activity - Following the tightening of U.S. export controls on AI chips to China, at least $1 billion worth of restricted Nvidia advanced AI processors have been shipped to mainland China [1]. - The B200 GPU has become the most popular chip in China's semiconductor black market, widely used by major U.S. companies like OpenAI, Google, and Meta for training AI systems [1][2]. - Despite the ban on selling advanced AI chips to China, it is legal for Chinese entities to receive and sell these chips as long as they pay the relevant border tariffs [1][2]. Group 2: Distribution and Sales Channels - A company named "Gate of the Era" has emerged as a major distributor of the B200, having sold nearly $400 million worth of these products [3]. - The B200 racks are sold at prices ranging from 3 million to 3.5 million RMB (approximately $489,000), which is lower than the initial price of over 4 million RMB [3]. - The sales of these chips are facilitated through various distributors in provinces like Guangdong, Zhejiang, and Anhui, with significant quantities being sold to data center providers [2][3]. Group 3: Market Dynamics and Future Outlook - The demand for Nvidia's B200 chips remains high due to their performance and relative ease of maintenance, despite U.S. export controls [11]. - Following the easing of the H20 export ban, the black market sales of B200 and other restricted Nvidia chips have reportedly decreased as companies weigh their options [13]. - Southeast Asian countries are becoming key transit points for Chinese companies to acquire restricted chips, with potential tightening of export controls being discussed by the U.S. government [13][15]. Group 4: Repair and Maintenance Services - There is a growing demand for repair services for Nvidia's high-end chips, with some companies in China specializing in the maintenance of H100 and A100 chips that have entered the market through special channels [17]. - The average monthly repair volume for these AI chips has reached 500 units, indicating a significant market need for maintenance services [17][18]. - The introduction of the H20 chip has seen limited market acceptance due to its high price and inability to meet the demands for training large language models [18].

Core Viewpoint - The "Winning the Race: America's AI Action Plan" aims to enhance U.S. competitiveness in AI technology against China, proposing over 90 policy recommendations to accelerate development in Silicon Valley [1] Group 1: Regulatory Changes - The plan marks a shift from prescriptive regulation to a more lenient governance approach, emphasizing the need to eliminate barriers to AI innovation and application [2] - It seeks to revoke state-level regulations perceived as burdensome, redirecting federal funding away from states with stringent AI regulations [3] Group 2: Support for Open Source AI - The plan promotes open innovation and public AI research infrastructure as key pillars for maintaining U.S. leadership in AI [4] - It provides strong government support for open-source AI models, encouraging small and medium enterprises to adopt these models [5] Group 3: Export Strategy - The plan emphasizes the importance of exporting U.S. AI technologies to meet global demand, with a directive to establish an AI export program within 90 days [6] - It aims to counter China's influence in the global AI market through enhanced export controls and strategic partnerships [7] Group 4: Ideological Stance - The plan lacks commitments to protect the public from AI-related harms but emphasizes the need for AI systems to align with American values and protect free speech [8] - It criticizes existing global AI initiatives for promoting regulations that conflict with U.S. values and highlights the perceived ideological bias in these efforts [8]

Group 1 - The core viewpoint of the article highlights a significant inflow of capital into the semiconductor equipment sector, with the semiconductor equipment ETF (159516) seeing a net inflow of 21 million units and over 120 million yuan in the past five days, indicating strong investor interest in semiconductor assets [1] - From a macro perspective, the domestic production and self-sufficiency of chips remain the main development line, despite temporary easing in US-China relations and Nvidia's H20 supply recovery, the potential risks of being constrained by key technologies should not be overlooked [1] - The semiconductor equipment ETF tracks the semiconductor materials and equipment index (931743), which includes listed companies involved in semiconductor materials and equipment manufacturing, reflecting the level of domestic semiconductor industry's self-sufficiency in core materials and equipment [1] Group 2 - The index components cover critical upstream areas of the semiconductor industry chain, including silicon wafers, photoresists, and etching machines, indicating a focus on enhancing the domestic production rate of computing chips [1] - Investors without stock accounts can consider the Guotai Zhongzheng Semiconductor Materials and Equipment Theme ETF Initiated Link C (019633) and A (019632) for exposure to this sector [1]

Group 1 - The core viewpoint emphasizes that intelligent agents represent a new evolutionary stage for large models, acting as a complement rather than a replacement [2][6][11] - The industry is currently divided into two main models for intelligent agents: one where large model vendors develop them, and another where application companies build on existing large models [2][8] - The domestic market faces challenges in monetizing intelligent agents due to high operational costs and a lack of established payment habits among users [8][19] Group 2 - Intelligent agents are expected to replace many low-level jobs, transforming employees into roles that define and manage these agents [14][16] - The future of intelligent agents is seen as a significant opportunity across various industries, with the potential to automate complex tasks and reduce reliance on human labor [14][16] - The concept of general intelligent agents is viewed skeptically, with a stronger belief in the rise of specialized intelligent agents tailored to specific industries [11][12][13] Group 3 - DeepSeek has contributed to the Chinese large model industry by eliminating redundant models and promoting an open-source ecosystem [18][19] - The decline in DeepSeek's traffic is acknowledged, but its foundational models continue to support many companies in the intelligent agent space [17][18] - The domestic chip industry is seen as having the potential to catch up with international competitors like NVIDIA, particularly in inference capabilities [19][20]

Group 1: Development of Large Models and Intelligent Agents - Current large models are limited and primarily function as chatbots, lacking true productivity capabilities [3] - Intelligent agents complement large models by executing complex tasks autonomously, enhancing enterprise applications [3] - Future intelligent agents should specialize in different industries rather than being generalized, akin to virtual consultants [3][4] Group 2: Business Model Challenges - The advertising model for AI, as seen with Manus, is not sustainable due to high operational costs and user demands [5] - Direct user charging may become necessary as AI task completion requires significantly more resources than traditional chat interactions [5] Group 3: Domestic Chip Procurement - The company is shifting towards domestic chip procurement, particularly from Huawei, despite acknowledging the performance gap with NVIDIA [6] - Emphasis is placed on the necessity of using domestic chips to drive improvement and innovation [6] Group 4: DeepSeek and Open Source Value - DeepSeek's recent traffic decline is attributed to its focus on AGI rather than app performance, with its true value lying in third-party applications [6] - The open-source nature of DeepSeek is seen as a strategic advantage against monopolistic practices [6][7] Group 5: Cybersecurity Risks - The deployment of large models introduces significant security risks, including hallucinations, lowered attack barriers, and advanced threats from hackers [8][9] - The company is developing solutions to counter these risks, including intelligent agents for real-time defense and monitoring [9] Group 6: AI Hardware Development - Upcoming AI hardware includes an AI recording pen and smart glasses, with the latter requiring a display to enhance functionality [10] - The company critiques the practicality of AI glasses without display features, emphasizing the need for diverse functionalities [10] Group 7: Commercialization of Intelligent Agents - There is a strong domestic market potential for intelligent agents, particularly for small and medium enterprises [10] - The company aims to lower costs and enable personalized intelligent agent creation for individual users [10]