Core Insights - The company reported a total revenue of RMB 44.02 billion for the first three quarters of 2023, representing a year-on-year growth of 36.8%, significantly surpassing last year's growth rate [1] - The gross profit reached RMB 16.21 billion, with a year-on-year increase of 20.1%, and a stable gross margin of 36.8% [1] - The company achieved its first quarterly profit in Q3 2023, driven by increasing demand for practical AI applications across various industries [2] Financial Performance - Revenue from the core product, the AI platform, surged to RMB 36.92 billion, marking a 70.1% year-on-year growth, contributing to the acceleration of overall company performance [2] - R&D expenses amounted to RMB 14.89 billion, an increase of 8.4% year-on-year, with an R&D expense ratio of 33.8%, down 8.9 percentage points, indicating improved efficiency in R&D investment [2] Strategic Investments - The company invested in domestic computing power by acquiring over 9% of a leading domestic GPU manufacturer and strategically investing in another [3] - The launch of ModelHub XC and the AI engine EngineX aims to enhance the synergy between domestic computing power and large models, facilitating the conversion of hardware capabilities into visible business value [3] Technological Advancements - The company introduced the Virtual VRAM expansion card, allowing single-card memory to be expanded to 256GB, supporting large-scale AI tasks without hardware replacement [3] - The company has developed GPU dynamic scheduling capabilities based on Kubernetes v1.34, significantly lowering the barriers to cloud-native computing management [4] Market Position and Future Outlook - The company has maintained its position as the market leader in China's machine learning platform sector for seven consecutive years, with plans to further develop its AI platform as a core engine for productivity across various industries [4] - The company is expanding into new fields such as AI combined with stablecoins and sports, with recent initiatives in AI-driven sports solutions aimed at enhancing training and promoting digital upgrades in fitness [4] Consumer Electronics Expansion - The Phancy consumer electronics business is gaining momentum, focusing on open technology collaborations to transition consumer electronics into consumer goods [5] - The company has launched AI smart glasses with a 13-megapixel camera and is collaborating with brands like AOC to innovate in smart wearable devices [5]

今天（11月12日），范式发布2025年前三季度业绩公告。公告显示，今年前三季度，公司总收入人民币 44.02亿元，同比增长36.8%，远超去年同期增速；毛利润人民币16.21亿元，同比增长20.1%，较去年同 期增速有明显提升；毛利率为36.8%，期间内总体保持平稳，特别值得一提的是，公司在第三季度首次 实现单季度盈利。 当然，这一强劲增长也离不开公司长期坚持在平台上探索并沉淀全栈式的AI技术及功能模块。本期间 研发费用为人民币14.89亿元，同比增加8.4%，研发费用率为33.8%，同比下降8.9个百分点，在保持技 术持续领先的同时，大幅优化研发投入产出效率。 解放算力潜能 加速应用端市场爆发 公告期间，公司深入投资发展国产算力——这一新时代的"石油"：一方面，公司投资持有国产领先GPU 厂商曦望超过9%的股份，也通过公司旗下基金战略投资了另一家国产领先GPU厂商天数智芯；另一方 随着企业客户对执行可落地、能够产生真实业务价值的AI应用需求愈发急迫，公司捕捉到大量新增业 务机会，并在能源、制造、金融、零售等多个关键行业拓展了新客户。公司核心产品先知AI平台业务 在公告期间实现爆发式增长，板块收入人民币36 ...

Overall Industry Performance - The electronic industry is experiencing a volatile and differentiated performance, with mixed results across various sectors [3] - The Shanghai Composite Index rose by 1.08%, while the Shenzhen Component Index increased by 0.19%, and the CSI 300 Index went up by 0.82% [2] Sector Analysis - The Other Electronics II sector showed the strongest performance with a weekly increase of 5.03%, leading the entire industry [3] - The Components sector rose by 2.14%, with printed circuit boards increasing by 1.90%, while passive components fell by 3.25%, indicating significant differentiation within the sector [3] - The Semiconductor sector remained flat overall, but there were notable internal differences, with semiconductor equipment rising by 4.80% and discrete devices increasing by 1.28%, while analog chip design and integrated circuit testing fell by 3.56% and 3.17% respectively [3] - The Optical and Optoelectronic sector slightly declined by 0.25%, with LEDs increasing by 1.07% and optical components decreasing by 1.04% [3] - The Consumer Electronics sector dropped by 2.45%, continuing its adjustment trend, with brand consumer electronics down by 6.22% [3] - The Electronic Chemicals II sector fell by 0.82%, with the Electronic Chemicals III sub-sector experiencing a similar decline [3] Notable Company Developments - Anshi China announced it has established sufficient inventory to meet customer demand until the end of the year [4] - Blue Arrow Electronics exited the semiconductor investment fund, terminating the co-investment matter [4] - Supermicro, in collaboration with Intel and Micron, set a world record in quantitative trading performance testing [4] - Samsung SDI is in talks with Tesla regarding LFP battery orders, targeting the U.S. energy storage market [4] - Lite-On Optoelectronics reported growth in both revenue and net profit for the first three quarters, with domestic OLED materials making significant breakthroughs [4] Investment Insights - The recent update of Cambrian's self-developed basic software platform NeuWare is seen as a comprehensive enhancement rather than a simple functional upgrade, shifting the value proposition of domestic computing power from "hardware replacement" to "platform replacement" [5] - As domestic hardware and software platforms accelerate their autonomy, the ecosystem for domestic chips may be strengthened, potentially leading to a new phase for domestic computing power [5]

Core Viewpoint - The short-term market is in a policy and performance vacuum, requiring more catalysts for the index to break through key levels, with the market likely to remain volatile [1] Group 1: Market Strategy - The recommendation is to maintain a "barbell" strategy in asset allocation [1] - In the technology sector, the pressure of crowding has eased, and after adjustments, the cost-effectiveness is gradually improving, focusing on low-position targets in areas such as Hang Seng Technology, domestic computing power, and AI applications [1] - Given the uncertainties both domestically and internationally, previously underperforming dividend stocks are experiencing a rebound, with current opportunities in banks and certain cyclical dividend stocks [1] Group 2: Sector Focus - The third-quarter reports and high-frequency indicators are cross-validating, indicating that varieties with non-crowded valuations and chips, such as new energy and chemicals, are worth attention [1] - In the medium term, seven key clues will be monitored: policy cycle, technology cycle, real estate cycle, capacity cycle, inventory cycle, energy cycle, and capital market reform, with advanced manufacturing and pro-cyclical consumption being potential decisive factors [1]

Core Viewpoint - The market is currently in a policy and performance vacuum, requiring more catalysts for the index to break through key levels, with a likely focus on volatility in the near term [1] Group 1: Market Strategy - The recommendation is to maintain a "barbell" strategy in asset allocation [1] - In the technology sector, the pressure from overcrowding has eased, and after adjustments, the cost-effectiveness is gradually improving, with a focus on low-position targets in areas such as Hang Seng Technology, domestic computing power, and AI applications [1] - Given the uncertainties both domestically and internationally, previously underperforming dividend stocks are expected to see a rebound, with current opportunities in banks and certain cyclical dividend stocks [1] Group 2: Sector Focus - The third-quarter reports and high-frequency indicators are cross-validating, indicating that varieties with non-crowded valuations and chips, such as new energy and chemicals, are worth attention [1] - In the medium term, seven key clues will be monitored: policy cycle, technology cycle, real estate cycle, capacity cycle, inventory cycle, energy cycle, and capital market reform, with advanced manufacturing and pro-cyclical consumption being potential decisive factors [1]

Group 1 - Alibaba Group is building large-scale AI infrastructure and increasing investment to create a super AI cloud [1] - Ant Group has deployed a domestic computing power cluster at the scale of ten thousand cards, achieving over 98% stability in training task performance [1] - Minsheng Securities believes that super nodes are expected to become mainstream technology under cluster expansion, enhancing the domestic computing power ecosystem [1] Group 2 - The industry is focusing on building ultra-high bandwidth and ultra-low latency ScaleUp networks to connect multiple XPU high-speed interconnections into super nodes [2] - Huaxi Securities suggests that regardless of future restrictions on overseas computing power chip sales to China, domestic computing power development will remain a key focus of policy support [2] Group 3 - Zhongke Shuguang has launched the world's first 640-card single-machine super node, which can support trillion-parameter large model training and improve performance by 30%-40% compared to traditional solutions [3] - Qihua Data is constructing high-end GPU computing server clusters and expanding its data storage business into the AI server field [3]

Investment Rating - The report maintains a "Recommendation" rating for the industry, indicating a positive outlook for the sector's fundamentals and expected performance relative to the benchmark index [3]. Core Insights - The report highlights the ongoing restrictions on NVIDIA's chip sales to China, which may impact the competitive landscape in the AI and computing sectors [6][9]. - The launch of the world's first single-cabinet 640-card super node, scaleX640, by Inspur is noted as a significant technological advancement, enhancing performance and efficiency in large-scale computing deployments [6][10]. - The domestic software companies are showing signs of recovery from a low base, with institutional holdings and expectations at low levels, suggesting potential for growth as AI applications evolve [6][17]. Summary by Sections Weekly Highlights - NVIDIA faces new restrictions on chip sales to China, affecting its market presence [9]. - Inspur has introduced the scaleX640 super node, which boasts significant performance improvements and is designed for large-scale deployments [10]. - The performance of overseas SaaS companies has been positively influenced by AI technology applications, with many reporting substantial revenue growth [12]. Market Performance Review - The computer sector experienced a decline of 2.54% in the first week of November 2025, with notable stock performances from companies like Chunzong Technology and Yingfang Software [18][19].

Investment Rating - The industry investment rating is maintained as "Positive" [4] Core Insights - The release of the Shuguang super node marks a significant breakthrough in domestic computing power, showcasing advantages over traditional intelligent computing methods through high-speed interconnection technology, which enhances bandwidth and reduces latency [6] - The Shuguang Scale X640 super node, launched on November 6, 2025, is the world's first 640-card single-cabinet super node, designed for training trillion-parameter models, improving performance by 30%-40% compared to traditional solutions [6] - Domestic computing power and super nodes are expected to continue achieving breakthroughs, with the Shuguang X640 demonstrating top-tier performance, efficiency, reliability, and openness in the AI computing system [6] Summary by Sections Investment Recommendations and Targets - Recommended stocks include Haiguang Information (688041, Buy), Zhongke Shuguang (603019, Buy), and others that are either rated or not rated [3] Industry Overview - The report focuses on the computer industry in China, highlighting advancements in domestic computing capabilities and the competitive landscape [4][6]

Investment Rating - The report assigns an "Increase" rating for the stocks mentioned, indicating a potential rise of over 15% relative to the CSI 300 index within the next 12 months [5][13]. Core Insights - The launch of the Shuguang scaleX640 marks a significant advancement in domestic computing power, achieving a 30-40% performance improvement in trillion-parameter model training and inference compared to traditional solutions [2][5]. - The scaleX640 enhances the cost-effectiveness of inference scenarios and is expected to accelerate the breakthrough of domestic computing chips in training applications [5]. - The report highlights the long-term stability of the scaleX640, which has undergone over 30 days of reliability testing, ensuring support for large-scale cluster deployments [5]. - The open architecture of the scaleX640 is anticipated to facilitate the integration of domestic computing software ecosystems and unify supernode structures, potentially leading to rapid iterations in domestic supernodes [5]. Summary by Sections Performance and Features - The scaleX640 boasts a 20-fold increase in computing density per cabinet and supports trillion-parameter model training, with a performance boost of 30-40% in inference scenarios compared to traditional methods [5]. - Innovative technologies such as ultra-high-speed orthogonal architecture, high-density blades, immersion phase change cooling, and high-voltage direct current power supply are employed in the scaleX640 [5]. Recommended Stocks - The report recommends the following stocks: - Cambrian-U (688256.SH) with a closing price of 1480.00 and an EPS forecast of 5.04 for 2025 [7]. - Haiguang Information (688041.SH) with a closing price of 239.68 and an EPS forecast of 1.34 for 2025 [7]. - SMIC (688981.SH) with a closing price of 124.85 and an EPS forecast of 0.46 for 2025 [7]. - Zhaoyi Innovation (603986.SH) with a closing price of 227.20 and an EPS forecast of 2.35 for 2025 [7]. - Shengke Communication-U (688702.SH) with a closing price of 119.31 and an EPS forecast of 0.04 for 2025 [7]. - The report also mentions Chipone Technology (688521) as a related stock [5].

Core Viewpoint - The article emphasizes the significant advancements in China's semiconductor industry, particularly in the context of domestic data centers and the increasing demand for local chip solutions, which may lead to a reduction in the market share of imported chips like NVIDIA's in China [2][4]. Group 1: Domestic Chip Advancements - The recent release of the world's first single-cabinet 640-card super node, scaleX640, by Zhongke Shuguang marks a major breakthrough in domestic chip technology, designed to compete directly with NVIDIA's CUDA ecosystem [4]. - The scaleX640 super node features a "one-to-two" high-density architecture, achieving ultra-high-speed interconnects and enabling the construction of large-scale, high-bandwidth, low-latency communication domains [4]. - Compared to similar products in the industry, the scaleX640 has doubled its overall computing performance and increased the computing density by 20 times within a single cabinet [4]. Group 2: Competitive Landscape - Despite some performance gaps in underlying chip capabilities compared to NVIDIA's NVL72 super node architecture, the scaleX640 has surpassed NVIDIA in cooling technology, power supply solutions, hardware architecture, computing integration, and power density [5]. - Huang Renxun, CEO of NVIDIA, stated that China is poised to win the AI race due to favorable regulatory environments and lower energy costs, while criticizing certain countries for hindering progress [5]. - The emergence of open computing clusters like scaleX640 is seen as a way to bridge the gap in computing power by fostering collaboration across the AI industry, positioning China's AI innovation as a critical component for future advancements [5].