证券研究报告·策略报告·策略深度报告 10 月度金股：聚焦高质量发展组合 2025 年 10 月 11 日 [Table_Summary] ◼ 10 月预计指数将保持震荡上行态势 假期海外市场与产业层面催化都偏正面。9 月市场受阅兵、节前避险、季 末调仓等因素影响走势偏震荡，而节后压制因素将逐步消退。此外，假期 期间全球主要股市表现突出，AI 产业催化密集，预计内地资金风险偏好也 有望提升。 预期四季度货币政策或将积极变化，推动权益市场上行。9 月底召开的央 行三季度例会调整了部分关键表述：二季度对宏观经济用"向好态势"， 本次改为"稳中有进"。结合美国降息周期与国内经济，四季度货币政策 加大刺激的必要性有所增强。若四季度政策/政策预期出现积极变化，在当 前市场情绪下，将对股市产生偏正面作用。 ◼ 结构上建议聚焦科技成长 从顶层设计思想出发，继续聚焦"高质量发展相关行业"。 9 月 30 日起， 《人民日报》连发 8 篇 "钟才文" 文章，核心是明确后续经济政策要保 障高质量发展态势，或为后续四中全会、"十五五"规划思路做铺垫。同 时，央行例会也侧重该领域，后续投资仓位应将继续向 "高质量发展相关 行业"靠拢 ...

21世纪经济报道记者 赵云帆 9月29日晚间，深度求索公司发布了 DeepSeek-V3.2-Exp 模型，其宛如再次向国产芯片江湖投入一枚巨 石。 官方公告显示，新版本模型实验性地引入了 DeepSeek Sparse Attention（一种稀疏注意力机制），其显 著降低计算资源消耗并提升推理效率。 而得益于这一技术创新，深度求索宣布其官方App、网页端、小程序均已同步更新为新模型，并全面下 调API价格，降幅在50%到75%不等。 值得注意的是，不同于上次发布 DeepSeek-V3.1版本时"犹抱琵琶半遮面"的状态，此次 V3.2 EXP 版本 的发布，几乎零时差引发数个国产芯片厂商——华为、寒武纪、海光等的"认领"。 而这也标志着，国产AI软硬件生态的协同，已经在 DeepSeek 横空出世之后初具雏形。 争相"认领" 就在 DeepSeek 新版放出之后仅仅4分钟，寒武纪就对新模型适配的国产芯片进行"率先认领"。 该公司宣布，已同步实现对深度求索公司最新模型 DeepSeek-V3.2-Exp 的适配，并开源大模型推理引擎 vLLM-MLU源代码。 紧随其后的则是华为昇腾，其表示，基于vLLM/S ...

Group 1 - The report from Dongfang Securities highlights that process technology upgrades drive chip performance improvements, while advanced packaging serves as another key driver for performance enhancement [1] - In the context of slowing process technology upgrades, increasing the die area can enhance transistor count and computing power, with Nvidia's H100 die area nearing the reticle limit of approximately 800-900 mm² [1] - Nvidia's B200 adopts advanced packaging with dual die integration, achieving 208 billion transistors in a single package, which is more than double the 80 billion transistors in the H100 [1] Group 2 - System assembly is emerging as a new driver for AI server performance enhancement, as wafer manufacturing and advanced packaging may not keep pace with the growing demand for AI computing power [2] - The number of GPUs in AI servers is expected to increase significantly, from 8 GPUs per server to 72 GPUs per cabinet, and potentially to 144 GPUs in the VR Ultra NVL576 cabinet by 2027 [2] - The increase in GPU count raises challenges in heat dissipation and system assembly complexity, with industry leaders likely to benefit from the elevated barriers to entry and improved competitive environment [2] Group 3 - In terms of investment targets, companies related to AI server system assembly are maintained, including Industrial Fulian, which has optimized GB200 series product testing and reduced system-level cabinet debugging time [3] - Industrial Fulian is expanding capacity globally and has introduced fully automated assembly lines, expecting strong growth in GB200 shipments, primarily driven by large North American cloud service providers [3] - Haiguang Information's merger with Zhongke Shuguang is anticipated to create vertical integration capabilities encompassing CPU, DCU, and system assembly [4] Group 4 - Lenovo is expected to launch various servers based on Blackwell Ultra starting in the second half of 2025, as indicated by Nvidia [4] - Huaqin Technology, a core ODM supplier for domestic internet companies' AI servers, is set to benefit from the capital expenditure expansion of downstream cloud companies [4]

Core Viewpoint - The article discusses the rapid development of the domestic AI chip industry in China, driven by policies promoting localization and self-control, and categorizes companies into three types based on their technology focus: ASIC manufacturers, CPU-focused companies, and those offering full-stack solutions [1][34]. Group 1: AI Chip Classification - AI chips can be categorized into cloud AI chips, edge AI chips, and terminal AI chips, with training and inference chips being the main types [3]. - The main types of AI chips include GPU, FPGA, and ASIC, with GPUs expected to hold 80% of the market share by 2025 according to IDC data [1][2]. Group 2: Performance Metrics - Key performance indicators for AI chips include computing power, power consumption, area, precision, and scalability, with computing power and power efficiency being critical metrics [4][5]. - The common units for measuring computing power are TOPS and TFLOPS, indicating the number of operations per second [4]. Group 3: Domestic AI Chip Landscape - The global AI chip market is dominated by NVIDIA, while domestic companies like Cambricon and Haiguang Information are emerging as significant players [7]. - A comparison of domestic AI chip companies reveals a variety of backgrounds, capitalizations, and funding rounds, indicating a diverse and competitive landscape [8]. Group 4: Company Profiles - Cambricon focuses on a complete product matrix for cloud and edge applications, utilizing a proprietary instruction set architecture optimized for deep learning tasks [10][11]. - Haiguang Information specializes in high-end processors, with its DCU series designed for AI acceleration, emphasizing compatibility with mainstream software [14][15]. - Other notable companies include Muxi Integrated Circuit, which targets high-performance GPGPU markets, and Tensu Intelligent Chip, which offers a self-developed general-purpose GPU [18][21]. Group 5: Technology and Innovation - Companies are adopting advanced manufacturing processes, with many using 7nm technology and some developing 5nm products [34]. - The article highlights the importance of software ecosystems and compatibility with mainstream AI frameworks to lower developer migration costs [37]. Group 6: Market Trends and Strategies - The domestic AI chip industry is focusing on performance improvement through multi-precision support and high-bandwidth memory optimization [37]. - Companies are increasingly collaborating with major AI models to enhance their chip offerings and ensure compatibility with existing software ecosystems [37].

Core Viewpoint - Haiguang Information has announced a stock incentive plan benefiting 879 employees, with a grant price significantly lower than the current market price, aimed at enhancing operational performance and aligning employee interests with company goals [2][4]. Group 1: Stock Incentive Plan Details - The stock incentive plan will grant up to 20.6843 million restricted shares at a price not lower than 90.25 yuan per share, which represents a discount of approximately 41% from the current price of 220.84 yuan [4]. - The plan targets 879 employees, accounting for about 31.36% of the total workforce of 2,803 as of June 30, 2025 [4]. - The source of the shares for the incentive plan is from the company's stock repurchase program, which began in 2023 at prices ranging from 55.00 yuan to 107.84 yuan per share [4]. Group 2: Performance Targets - The performance targets for the incentive plan are primarily based on revenue growth, with specific goals set for 2024 and the following years: a 55% increase for 2025, 125% for 2026, and 200% for 2027 [4][5]. - In 2024, Haiguang Information achieved a revenue growth rate of 52.40%, and for the first half of 2025, the growth rate was 45.21%, indicating that meeting the 2025 performance targets may be feasible [5]. Group 3: Exclusion of Merged Entity Revenue - The performance assessment will not include revenue from the ongoing merger with Zhongke Shuguang, allowing for a more focused evaluation of Haiguang Information's core business performance [3][6]. - The decision to exclude the merged entity's revenue is intended to ensure that the incentive plan accurately reflects the contributions of the employees and avoids dilution of the company's own growth metrics [9].

Core Viewpoint - The digital economy theme index and related stocks are experiencing significant growth, driven by strong demand in the AI and digital infrastructure sectors, indicating a positive outlook for the industry. Group 1: Digital Economy Index Performance - As of September 11, 2025, the CSI Digital Economy Theme Index (931582) rose by 2.25%, with notable increases in constituent stocks such as Haiguang Information (688041) up by 9.87% and Cambricon (688256) up by 6.28% [1] - The Digital Economy ETF (560800) increased by 2.16%, with a latest price of 0.95 yuan and a turnover rate of 1.28%, totaling 9.4021 million yuan in trading volume [1] - Over the past month, the average daily trading volume of the Digital Economy ETF reached 33.1933 million yuan [1] Group 2: Company Performance - Haiguang Information reported a revenue of 5.46 billion yuan for H1 2025, representing a year-over-year increase of 45.2%, with a net profit of 1.2 billion yuan, up by 40.8% [1] - In 2024, Haiguang Information achieved a revenue of 9.16 billion yuan, a year-over-year growth of 52.4%, and a net profit of 1.93 billion yuan, up by 52.9% [1] - The gross margin for Haiguang Information was 60.1%, showing a slight decrease compared to the previous year [1] Group 3: AI Market Outlook - According to IDC, the Chinese AI market is projected to reach 815.9 billion yuan by 2028, with a five-year compound annual growth rate (CAGR) of approximately 33% [2] - The AI industry is characterized by a gradient of development: upstream infrastructure expansion (GPU/memory/network), midstream model commercialization, and downstream application proliferation, particularly in enterprise-level AI agents [2] - The digital economy theme index includes companies with high digitalization levels and infrastructure relevance, reflecting the overall performance of the digital economy sector [2] Group 4: Top Constituents of the Index - As of August 29, 2025, the top ten weighted stocks in the CSI Digital Economy Theme Index accounted for 53.36% of the index, including Dongfang Wealth (300059), Cambricon (688256), and Haiguang Information (688041) [2]

Core Viewpoint - The domestic AI chip industry is experiencing significant growth driven by strong demand for AI inference, with leading companies accumulating inventory and securing key materials to prepare for future market developments [2][3]. Financial Performance - Haiguang Information achieved a revenue of 5.464 billion yuan in the first half of the year, a year-on-year increase of 45.21%, with a net profit of 1.201 billion yuan, up 40.78% [3]. - Cambrian's revenue surged to 2.881 billion yuan, marking a staggering year-on-year growth of 4347.82%, with a net profit of 1.038 billion yuan, compared to a loss of 530 million yuan in the previous year [3]. - Longxin Zhongke reported a revenue of 244 million yuan, a 10.9% increase year-on-year, but still faced a net loss of 294 million yuan, which is a 23.53% decline compared to the previous year [3]. Inventory and Contract Liabilities - Cambrian's inventory reached 2.69 billion yuan, accounting for 31.95% of total assets, with a 51.64% increase from the previous year [5]. - Cambrian's contract liabilities grew to 543 million yuan, representing 6.45% of total assets, a dramatic increase of 61223.22% year-on-year [5]. - Haiguang Information's contract liabilities rose by 242.1% to 3.091 billion yuan, making up 9.57% of total assets, driven by customer prepayments [5]. Market Dynamics - Domestic AI chip manufacturers are actively increasing inventory as part of strategic development considerations, with a focus on key raw materials like HBM and wafers [6][7]. - The domestic AI chip market is characterized by diverse technology routes, with companies like Haiguang Information and Cambrian competing in the GPU space, while others like Huawei HiSilicon and Cambrian focus on ASIC custom chips [7][8]. Commercialization Progress - Domestic AI chip companies have made notable strides in commercialization, with Huawei HiSilicon holding approximately 23% market share in the AI acceleration chip market in China [8]. - Cambrian's products are being deployed in key industries such as telecommunications, finance, and the internet, while Haiguang's CPU and DCU series are widely applicable in big data processing and AI [8][9]. Ecosystem Development - The development of a collaborative and open ecosystem is crucial for the sustained growth of the domestic AI chip industry, with companies like Haiguang Information emphasizing the need for deep collaboration across the supply chain [11][12]. - Increased marketing expenditures by Haiguang Information, which rose by 185.83% to 203.4 million yuan, reflect efforts to expand market presence and enhance ecosystem development [12].

Investment Rating - The report maintains a "Buy" rating for the company, with a target price based on the last closing price of 201.89 [1][6]. Core Insights - The company has demonstrated sustained high growth, with a focus on building a comprehensive computing power industry chain [6][12]. - In H1 2025, the company achieved operating revenue of 5.464 billion, a year-on-year increase of 45.21%, and a net profit attributable to shareholders of 1.201 billion, up 40.78% [4][5]. - The company is expected to benefit from the trend of domestic computing power localization and the rapid growth of AI technology, with predictions indicating that China's intelligent computing power scale will reach 2,781.9 EFLOPS by 2028, with a compound annual growth rate of 46.2% [6][8]. - The merger with Zhongke Shuguang aims to create a vertically integrated capability from chip design to server manufacturing and cloud computing services, enhancing competitiveness against international leaders [6][8]. Financial Performance - In Q2 2025, the company reported revenue of 3.064 billion, a year-on-year increase of 41.15%, and a net profit of 696 million, up 23.14% [5]. - The gross margin for Q2 was 59.33%, reflecting a decrease of 4.5 percentage points due to product mix adjustments and increased material costs [5]. - The company’s operating cash flow reached 2.177 billion in H1 2025, compared to a negative 113 million in the same period last year, indicating strong cash generation [5][6]. Revenue and Profit Forecast - The company is projected to achieve operating revenues of 13.949 billion, 19.906 billion, and 26.938 billion for 2025, 2026, and 2027 respectively, with net profits of 3.028 billion, 4.509 billion, and 6.324 billion [6][8]. - The expected revenue growth rates are 52.25% for 2025, 42.70% for 2026, and 35.33% for 2027 [8][14].

Core Viewpoint - The article highlights the selection of 240 stocks as "golden stocks" by brokerages, with a focus on those recommended by multiple firms, indicating strong investment interest in these companies [1] Group 1: Company Summaries - **Luoyang Molybdenum (603993.SH)**: A leading global producer of copper and cobalt, benefiting from rising prices and increased production, with a net profit growth of 55.49% in H1 2025 [2] - **Muyuan Foods (002714.SZ)**: A top player in pig farming with a significant cost advantage, experiencing a 952.92% increase in net profit in H1 2025 due to rising pig prices [3][4] - **AVIC Shenyang Aircraft (600760.SH)**: A core manufacturer of fighter jets, facing a decline in revenue and profit in H1 2025 but with strong future order potential [5] - **ZTE Corporation (000063.SZ)**: The fourth-largest global telecom equipment provider, seeing a 54.39% stock price increase driven by AI demand, despite a decline in net profit [6] - **Kingsoft Office (688111.SH)**: A leading office software provider with a strong user base, experiencing growth in subscription revenue but facing high valuation concerns [9] - **NewEase (300502.SZ)**: Specializes in optical modules with a significant market share, achieving a remarkable 340.13% stock price increase due to AI demand [10] - **Haiguang Information (688041.SH)**: Develops high-end processors, benefiting from AI demand and a strong order backlog, but facing high valuation risks [11] - **Luzhou Laojiao (000568.SZ)**: A leading producer of strong-flavor liquor, experiencing a 20.50% stock price increase despite a decline in revenue and profit [13] - **Shede Spirits (600702.SH)**: A liquor company with a diverse product range, seeing a 31.25% stock price increase amid expectations of consumption recovery [15] - **Zhaoyi Innovation (603986.SH)**: A semiconductor company benefiting from rising storage chip prices, with a 56.16% stock price increase and a forecasted profit growth of 41.52% [17] Group 2: Market Trends and Investment Logic - The overall market sentiment is positive for the selected stocks, driven by sector-specific demand and price increases, particularly in commodities and technology [1][2][3][4][5][6][10][11][13][15][17] - The companies are positioned well within their respective industries, with strong competitive advantages and growth potential, although some face high valuations and market risks [9][10][11][13][15][17]

Group 1: Market Overview - The A-share index rose by 1.45% on August 22, surpassing 3800 points, with significant growth in computing power and chip stocks [1] - The domestic computing power sector is experiencing a surge in activity, with multiple catalysts emerging [1] - Domestic AI chip manufacturers are gradually gaining market share, with Huawei's Ascend chips and Cambricon's AI chips making notable advancements [2][12] Group 2: Competitive Landscape - Nvidia dominates the AI acceleration chip market with a 70% share, while Huawei's Ascend holds 23% in 2024 [2] - The GPU market is primarily led by Nvidia, AMD, and Intel, with Nvidia expected to hold a 92% market share in independent graphics cards by Q1 2025 [5] - Domestic manufacturers are closing the gap in graphics rendering GPUs, with companies like Chipone and Jingjia Micro making significant progress [9][11] Group 3: Technological Advancements - Nvidia's GPU architecture has been updated approximately every two years since 2006, with the latest GeForce RTX 40 series showcasing significant performance improvements [8] - Domestic firms are innovating in ASIC technology, with companies like Cambrian and Horizon achieving notable advancements in performance and efficiency [12] - The CUDA software ecosystem developed by Nvidia is a key factor in its market leadership, providing a robust platform for AI training and deep learning applications [13][14] Group 4: Regulatory Environment - The U.S. has intensified export controls on AI chips to China, affecting high-end products from Nvidia and AMD [15] - New regulations have expanded the scope of restricted products, impacting several Nvidia models, including the H20, which was also placed under export control [15]