截至月份: 2025年10月31日 狀態: 新提交 致：香港交易及結算所有限公司 公司名稱: 北京第四範式智能技術股份有限公司 (於中華人民共和國註冊成立的股份有限公司) FF301 股份發行人及根據《上市規則》第十九B章上市的香港預託證券發行人的證券變動月報表 呈交日期: 2025年11月4日 I. 法定/註冊股本變動 | 1. 股份分類 | 普通股 | 股份類別 | H | | 於香港聯交所上市 (註1) | | 是 | | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 證券代號 (如上市) | 06682 | 說明 | | | | | | | | | | 法定/註冊股份數目 | | | 面值 | | 法定/註冊股本 | | | 上月底結存 | | | 320,809,496 | RMB | | 1 RMB | | 320,809,496 | | 增加 / 減少 (-) | | | 0 | | | RMB | | 0 | | 本月底結存 | | | 320,809,496 | RMB | | 1 RMB | | 320,809,496 ...

Core Insights - The "OpenMLDB: Key Technologies and System Platform for Enterprise-level Machine Learning Database" project, developed by 4Paradigm in collaboration with Shanghai Jiao Tong University and Tsinghua University, won the CCF Technology Achievement Award for its significant breakthroughs in enterprise-level machine learning data management [1][5]. Group 1: Technological Breakthroughs - The OpenMLDB project addresses significant challenges in data management for AI applications, particularly in core business systems, focusing on data consistency, real-time processing, and efficient preparation of large training datasets [2]. - Innovations include achieving data calculation consistency between online inference and offline training, enhancing the accuracy and reliability of AI applications [2]. - The project has developed millisecond-level real-time data processing technology, supporting immediate responses in critical areas such as financial risk control and intelligent recommendations [2]. - It has also improved the efficiency of batch preparation for large-scale training data, significantly shortening feature engineering and model iteration cycles [2][3]. Group 2: Economic Impact - The OpenMLDB project has been successfully deployed in various industries, including finance and e-commerce, demonstrating its capability to enhance system stability, data accuracy, and processing performance [4]. - In a leading bank, the project enabled millisecond-level real-time feature computation, effectively reducing financial losses from fraudulent transactions [4]. - The project supports the online recommendation system of Vipshop, improving recommendation accuracy and user experience, which directly contributes to an increase in GMV [4]. - The successful application of OpenMLDB has led to significant efficiency improvements and cost optimization for its users, providing a reusable technical pathway for intelligent transformation in related industries [4]. Group 3: Open Source and Innovation - 4Paradigm actively operates and builds the OpenMLDB open-source community, promoting the open and collaborative development of technology [5]. - This initiative accelerates the dissemination and popularization of project results, gathering global developer insights to advance enterprise-level machine learning data management technology [5]. - OpenMLDB has established itself as a domestically developed open-source solution with independent intellectual property rights, lowering the barriers for enterprises and developers to adopt advanced AI data management technologies [5]. - The CCF Technology Achievement Award recognizes significant discoveries and innovations in computer science and technology, with the OpenMLDB project standing out among 94 submissions [5].

FF301 股份發行人及根據《上市規則》第十九B章上市的香港預託證券發行人的證券變動月報表 截至月份: 2025年9月30日 狀態: 新提交 致：香港交易及結算所有限公司 公司名稱: 北京第四範式智能技術股份有限公司 (於中華人民共和國註冊成立的股份有限公司) 呈交日期: 2025年10月2日 I. 法定/註冊股本變動 | 1. 股份分類 | 普通股 | 股份類別 | H | | | 於香港聯交所上市 (註1) | | 是 | | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 證券代號 (如上市) | 06682 | 說明 | | | | | | | | | | | 法定/註冊股份數目 | | | 面值 | | | 法定/註冊股本 | | | 上月底結存 | | | 320,809,496 | RMB | | | 1 RMB | | 320,809,496 | | 增加 / 減少 (-) | | | 0 | | | | RMB | | 0 | | 本月底結存 | | | 320,809,496 | RMB | | | 1 RMB ...

Investment Rating - The report maintains a "Recommendation" rating for the computer industry, expecting the industry index to rise more than 5% over the next 3-6 months compared to the benchmark index [15]. Core Insights - The release of DeepSeek V3.2-Exp model is expected to drive the development of the domestic AI ecosystem, showcasing the synergy between domestic chips and large models, which leads to significant improvements in computational efficiency and cost reduction [5]. - The introduction of the DeepSeek Sparse Attention mechanism allows for a substantial enhancement in long text processing efficiency, overcoming traditional computational complexity limitations [5]. - The cost of AI applications is anticipated to decrease significantly, with the DeepSeek API prices reduced by over 50%, making advanced AI models more accessible to developers and businesses [5]. - Investment focus is recommended on domestic chip manufacturers like Cambricon and Hygon, as well as companies within the Huawei supply chain and various AI application firms [5]. Industry Basic Data - The computer industry comprises 337 listed companies with a total market capitalization of approximately 61,619.92 billion and a circulating market capitalization of about 55,710.57 billion [2]. Relative Index Performance - The absolute performance of the computer industry over the past 1 month, 6 months, and 12 months is -1.6%, 24.3%, and 72.1% respectively, while the relative performance is -4.3%, 6.3%, and 47.3% [3].

Core Viewpoint - The Hong Kong stock market's AI concept stocks are experiencing a collective surge, with Fourth Paradigm (6682.HK) rising nearly 8% to HKD 68.5, marking its second consecutive day of gains and reaching a new high since March 2024 [1] Industry Summary - DeepSeek announced the update of its official App, web version, and mini-program to DeepSeek-V3.2-Exp, significantly reducing service costs due to new model implementations, with API prices dropping by over 50% [1] Company Summary - Fourth Paradigm has launched the "Virtual VRAM" plug-in virtual memory expansion card, which converts physical memory into a dynamically scheduled memory buffer pool, allowing for elastic expansion of GPU computing resources [1] - The "Virtual VRAM" creates a high-speed data channel between memory and video memory, effectively increasing the virtual memory capacity of a single graphics card to a maximum of 256GB without major hardware changes [1]

Core Insights - The company reported a total revenue of 2.626 billion yuan for the first half of 2025, representing a year-on-year growth of 40.7% [2] - The core business, "Fourth Paradigm Prophet AI Platform," generated significant revenue of 2.149 billion yuan, with a year-on-year increase of 71.9%, accounting for 81.8% of total revenue [2] - The SHIFT intelligent solutions generated revenue of 371 million yuan, making up 14.1% of total revenue, focusing on key industries and strategic customer applications [2] - The company reported an operating loss of 848.8 million yuan, a reduction in loss of 55.1% compared to the same period last year [2] - The total assets of the company as of June 30, 2025, were 7.759 billion yuan, a growth of 2.26% from December 31, 2024 [3] - Total liabilities decreased to 1.497 billion yuan, down 39.78% from December 31, 2024 [3]

Core Viewpoint - The AI concept stocks in the Hong Kong market showed a positive trend, with most stocks experiencing gains on September 30th [1] Group 1: Stock Performance - Huilyang Technology (01860.HK) increased by 4.47%, reaching HKD 19.88 [1] - Maifushi (02556.HK) rose by 4.33%, trading at HKD 51.35 [1] - Chuangxin Qizhi (02121.HK) saw a gain of 3.65%, priced at HKD 7.95 [1] - Fourth Paradigm (06682.HK) climbed by 3.15%, with a share price of HKD 65.5 [1] - Meitu Inc. (01357.HK) increased by 3.26%, now at HKD 9.16 [1]

Group 1 - AI concept stocks saw a majority increase in early trading, with notable gains from companies such as 汇量科技 (4.47% increase), 迈富时 (4.33% increase), and 创新奇智 (3.65% increase) [1] - DeepSeek announced a significant update to its services, reducing the cost of its API by over 50% due to a new model that lowers service costs [1] - The National Development and Reform Commission (NDRC) plans to support various enterprises, including private companies, to deeply engage in AI initiatives [1] Group 2 - Huatai Securities indicated that the reduction in model prices will attract more developers to create AI applications, potentially boosting demand for computing power and increasing the likelihood of Super Apps [2] - Zhongyin International believes that the commercialization inflection point for AI applications is approaching, driven by improvements in reasoning efficiency and cost-effectiveness of domestic chips [2] - The combination of advancements in model capabilities, data accumulation, and synthetic data technology is facilitating a shift from "single-point breakthroughs" to "systematic collaboration" in AI capabilities, paving the way for large-scale commercialization [2]

Core Insights - The rapid development of AI large models has highlighted GPU memory capacity as a critical bottleneck for model training and inference efficiency [1] - Fourth Paradigm has launched the "Virtual VRAM" plug-in virtual memory expansion card, which transforms physical memory into a dynamically scheduled memory buffer pool, allowing for elastic expansion of GPU computing resources [1][2] - The new product aims to address the high costs associated with traditional GPU memory expansion and improve the scalability and multi-tasking capabilities of AI models [1][3] Company Overview - Fourth Paradigm's "Virtual VRAM" can expand the virtual memory capacity of a single GPU card up to 256GB, significantly enhancing the performance of existing NVIDIA GPUs without requiring hardware changes [2] - The product targets two main application scenarios: alleviating memory shortages during single-card operations and enabling multiple models to be deployed on the same GPU in low-load situations [2] Industry Implications - As the number and parameter scale of AI models continue to grow rapidly, memory capacity has become a key factor in building AI capabilities and controlling costs for enterprises [3] - The introduction of "Virtual VRAM" is expected to provide a cost-effective computing expansion solution, helping users maintain high performance while reducing costs [3] - Fourth Paradigm plans to collaborate with more memory manufacturers to further optimize and popularize AI infrastructure [3]

Core Insights - The rapid development of AI large models has highlighted GPU memory capacity as a critical bottleneck for training and inference efficiency [1][3] - Fourth Paradigm has launched the "Virtual VRAM" plug-in virtual memory expansion card, which transforms physical memory into a dynamically scheduled memory buffer pool, allowing for elastic expansion of GPU computing resources [1][2] Company Overview - Fourth Paradigm's "Virtual VRAM" can expand the virtual memory capacity of a single graphics card up to 256GB, significantly enhancing the capabilities of existing GPUs without requiring hardware changes [2] - The product is designed for two main application scenarios: addressing insufficient memory for large model single-card operations and enabling multiple models to be deployed on the same GPU in light-load scenarios [2] Industry Implications - As the number and parameter scale of AI models continue to grow rapidly, memory capacity has become a key factor in building AI capabilities and controlling costs for enterprises [3] - The new product from Fourth Paradigm is expected to provide a cost-effective computing expansion solution, helping users maintain high performance while achieving cost reduction and efficiency improvement [3] - Future plans include collaborations with more memory manufacturers to further optimize and popularize AI infrastructure [3]