[Table_Main] 公司研究|互联网行业|短视频行业 证券研究报告 快手(1024.HK)点评报告 2025 年 8 月 22 日 线上营销业务，25Q2 实现收入 197.65 亿元，同比+12.8%，借助 AI 大模 型技术优化智能营销产品解决方案，公司单 DAU 线上营销收入为 92.4 元。 公司借助大语言模型和模型序列建模，提升用户线索转化成功率，25Q2 UAX 解决方案渗透率持续提升，消耗占比外循环达 65%；内循环，公司迭 代智能出价等工具，商家投放素材稳定性提升。电商业务，25H1 GMV 为 6912.04 亿元，同比+16.5%，Q2 月活买家用户数达 1.34 亿，25Q2，含电 商在内的其他服务收入为 52.37 亿元，同比+26.0%，直播业务实现收入 100.44 亿元，同比+7.98%。25Q2，公司海外收入 13 亿元，同比+20.5%。 投资建议与盈利预测 公司为行业领先的内容社区和社交平台，持续推进 AI 战略，可灵商业化进 展超预期，利润有望持续释放。我们预计公司 2025-2027 年经调整净利润 200.77/246.20/287.21 亿元，维持公司"买入" ...

证券研究报告 快手-W (1024 HK) 港股通 2Q 业绩超预期，可灵 B 端商业化在即 2025 年 8 月 22 日│中国香港 互联网 快手 2Q 业绩：营收同增 13%至 350 亿，高于 VA 一致预期 2% (下同)。毛 利率同比改善 0.3pct 至 55.7%，经调整净利润同增 20%至 56 亿（超预期 11%）。分业务看，直播、广告、电商收入同比分别变动 8%、13%、26%， 广告方面，我们预期 3Q 广告同比增长 13%，弱于一致预期 16%，主系海 外广告受监管影响（广告收入占比中位数）；当前货架场景广告位正加速商 业化，内循环广告货币化率有望持续提升。我们预期可灵 25 年收入或达 10 亿+，2B 订阅方案有望于 3Q 推出，推动 B 端商业化加速。我们持续看好快 手在 AI 应用场景的长期渗透率提升。维持"买入"评级。 可灵商业化加速落地，新品"可灵画布"驱动 B 端渗透率 2Q 可灵收入 2.5 亿元，推理层面毛利率已转正。管理层指引 25 年收入为 1.25 亿美元，Capex 指引从 100 亿上提至 120 亿（去年 80 亿）。我们认 为可灵 2B 端变现加速，专业 ...

Financial Data and Key Metrics Changes - Total revenue increased by 13.1% year over year to RMB 35 billion in Q2 2025, with adjusted net profit reaching RMB 5.6 billion and an adjusted margin of 16% [5][34][40] - Gross profit grew by 13.8% year over year to RMB 19.5 billion, with a gross profit margin of 55.7%, reflecting a 0.4 percentage point increase year over year [38] - Selling and marketing expenses rose by 4.6% year over year to RMB 10.5 billion, accounting for 30% of total revenue, down from 32.4% in Q2 last year [39] Business Line Data and Key Metrics Changes - Revenue from online marketing services increased by 12.8% year over year to RMB 19.8 billion, driven by enhanced AI capabilities [35][56] - E-commerce GMV rose by 17.6% year over year, with the number of monthly average paying users reaching 134 million [19][74] - Revenue from Clean AI surpassed RMB 250 million in Q2, indicating strong growth in AI-driven services [9][36] Market Data and Key Metrics Changes - Average DAUs on the Kuaishou app reached an all-time high of 409 million, with MAUs at 715 million, reflecting a year-over-year increase of 3.4% and 3.3% respectively [11][12] - Revenue from local services grew by 20% year over year to RMB 1.3 billion, with DAUs in Brazil showing stable growth [28][29] Company Strategy and Development Direction - The company is focused on leveraging AI technology to enhance user experience and operational efficiency across its business lines [4][34] - Clean AI is positioned as a one-stop creative engine, with plans to expand its applications in gaming and professional film production [8][49] - The company aims to maintain high-quality growth while exploring synergies between content and business ecosystems [32][43] Management's Comments on Operating Environment and Future Outlook - Management expressed confidence in the company's long-term growth prospects, highlighting the resilience of its business ecosystem amid macro uncertainties [5][6] - The company plans to continue investing in AI technologies to drive efficiency and discover new commercialization opportunities [32][84] Other Important Information - A special dividend of HKD 0.46 per share was declared for the first time since delisting, amounting to approximately HKD 2 billion in total [5][6] - The company will no longer disclose quarterly GMV figures separately starting in 2026, focusing instead on a more comprehensive set of indicators [41][42] Q&A Session Summary Question: What are the major use cases for Clean AI users at the moment? - Clean AI's users include mass creators and professional creators, with applications in content creation, advertising, and film production [46][47] Question: What are the AI use cases in the overall business? - AI technology is integrated across various business scenarios, enhancing marketing material generation and improving user engagement [54][56] Question: What verticals are expected to have strong growth in the second half of the year? - Strong growth is anticipated in local services, automotive, and content consumption industries, with strategies to enhance client outreach and operational results [62][64] Question: How does Kuaishou maintain momentum in e-commerce amid competition? - The company focuses on synergizing supply and demand, enhancing merchant capabilities, and improving user engagement to drive e-commerce growth [72][74] Question: What updates are there on AI CapEx and net profit margin? - AI-related CapEx is expected to double, with the impact on overall profitability projected to remain around 1% to 2% [82][84]

Financial Data and Key Metrics Changes - Total revenue increased by 13.1% year over year to RMB 35 billion in Q2 2025, with adjusted net profit rising by 20.1% to RMB 5.6 billion, achieving a margin of 16% [7][37][38] - Gross profit grew by 13.8% year over year to RMB 19.5 billion, with a gross profit margin of 55.7%, reflecting a 0.4 percentage point increase year over year [40][41] - Selling and marketing expenses rose by 4.6% year over year to RMB 10.5 billion, accounting for 30% of total revenue, down from 32.4% in the previous year [42] Business Line Data and Key Metrics Changes - Revenue from online marketing services reached RMB 19.8 billion, up 12.8% year over year, driven by enhanced AI capabilities [38][39] - E-commerce GMV rose by 17.6% year over year, with the number of average monthly paying users reaching 134 million [22][23] - Revenue from Clean AI surpassed RMB 250 million, indicating strong growth in AI-driven services [12][39] Market Data and Key Metrics Changes - Average DAUs on the Kuaishou app reached an all-time high of 409 million, with MAUs at 715 million, reflecting a year-over-year increase of 3.4% and 3.3% respectively [14][6] - Revenue from external marketing services continued to grow, driven by strong demand from content consumption, local services, and automotive industries [19][20] Company Strategy and Development Direction - The company is focused on integrating AI technology across its business, enhancing user experience, and optimizing marketing solutions [36][37] - The strategy includes expanding Clean AI's applications in gaming and professional film production, aiming to empower creators and enhance operational efficiency [51][53] - The company plans to discontinue separate GMV disclosures starting in 2026, focusing on a more nuanced combination of performance indicators [44][45] Management Comments on Operating Environment and Future Outlook - Management expressed confidence in the company's long-term growth prospects, emphasizing the resilience of its business ecosystem amid macro uncertainties [7][8] - The company aims to maintain high-quality growth while exploring new commercialization opportunities through AI [35][46] - Future investments will continue to focus on AI technology to enhance operational efficiencies and drive sustainable growth [86][87] Other Important Information - A special dividend of HKD 0.46 per share was declared for the first time since delisting, totaling approximately HKD 2 billion [8] - The company has repurchased shares amounting to HKD 1.9 billion, representing about 0.9% of total shares outstanding [43] Q&A Session Summary Question: What are the major use cases for Clean AI users at the moment? - Clean AI's users include mass creators and professional creators, with applications in content creation, advertising, and film production [48][50] Question: What are the AI use cases in the overall business? - AI technology is integrated across various business scenarios, enhancing marketing material generation and improving user engagement [57][59] Question: What verticals are expected to have strong growth in the second half of the year? - Growth is anticipated in local services, automotive, and content consumption industries, with strategies to enhance client outreach and marketing efficiency [66][68] Question: How does Kuaishou maintain momentum in e-commerce amid competition? - The company leverages a synergized ecosystem and tailored initiatives for merchants, focusing on user acquisition and repeat purchases [74][76]

Core Viewpoint - Kuaishou's performance during the 618 shopping festival indicates strong growth in its general merchandise volume (GMV), highlighting the platform's potential for commercial monetization through innovative advertising strategies and enhanced user engagement via new technologies [2][3][4]. Group 1: E-commerce Performance - During the 618 period, Kuaishou's general merchandise card GMV increased by over 53% year-on-year, search GMV surged by over 143%, and short video GMV rose by over 29%, indicating a robust growth trajectory that outpaces the overall market [3]. - The general merchandise sector is becoming a crucial channel for users to browse, discover, and purchase products, with clearer pathways between content consumption and product conversion [3]. Group 2: Advertising Commercialization Potential - Kuaishou's general merchandise sector has significant potential for further advertising commercialization, expected to contribute additional revenue in the second half of 2025 [3]. - The platform is focusing on enhancing advertising efficiency for small and medium-sized merchants through traffic distribution, supply chain support, and intelligent tools, aiming to establish a conversion chain for advertising in the general merchandise context [3]. Group 3: Technological Advancements - The newly launched end-to-end generative recommendation system, OneRec, is anticipated to improve user engagement metrics such as time spent on the platform and user retention [4]. - OneRec utilizes a multi-modal AI model framework to enhance content understanding and recommendation accuracy, achieving a tenfold increase in effective computational capacity and reducing operational costs to 10.6% of traditional solutions [4]. - After its implementation, Kuaishou App and its Lite version saw an increase in user stay duration by 0.54% and 1.24%, respectively, with a significant growth in the 7-day user lifecycle [4]. Group 4: Content Innovation - Kuaishou collaborated with its self-developed AI model, Keling, to produce the AIGC series "New World Loading," which features all scenes generated by AI and includes various styles such as realism, science fiction, and animation [5]. - The first episode, released on June 26, 2025, achieved over 55 million views on Kuaishou Lite by June 30, showcasing the platform's capability in content generation through advanced technology [5]. - The segment "Martin Syndrome" from the series won the "Best Technology Award" at the 15th Beijing International Film Festival, reflecting Keling's technical prowess in AIGC production [5].

Macro Insights - The manufacturing PMI index continued to rise in June, driven by a reduction in external disturbances and a slight improvement in new export orders [2] - The internal economic momentum is recovering, with high-energy-consuming industries stabilizing and new economic drivers expanding steadily, leading to increased production and procurement activities [2] - The service industry business activity index saw a slight decline due to the end of holiday effects, but overall market expectations remain positive with the upcoming summer consumption peak [2] - The construction industry business activity index stabilized and improved, primarily due to positive developments in housing construction activities [2] Stock Recommendations - The A-share stock selection for July includes New Guodu, Hengsheng Electronics, Gree Electric Appliances, Haier Smart Home, Akol, New China Life, China Life, Dongfang Fortune, China National Materials, and Huayou Cobalt [3] - The Hong Kong stock selection for July includes Hong Kong Exchanges and Clearing, AIA, China Hongqiao, Tencent Holdings, Xiaomi Group-W, Xindong Company, Pop Mart, and Hua Hong Semiconductor [3] Company Research - Akol's COC project has passed the acceptance review and has officially entered the stable production phase, with successful output of qualified products [4] - Kuaishou's advertising growth is expected to accelerate in the second half of 2025, driven by the launch of the new recommendation system OneRec and the potential for further monetization of the general merchandise shelf [5] - Bosideng achieved a revenue of 25.9 billion yuan for the fiscal year 2024/2025, a year-on-year increase of 11.6%, with a net profit of 3.51 billion yuan, up 14.3% [6] - Health元 is transitioning its traditional main business and is expected to see orderly progress in innovation, with revised net profit forecasts for 2025 and 2026 [8] - New Dairy is expected to benefit from the continued advantages of raw milk and structural upgrades that enhance profitability, despite some concerns about the sustainability of raw milk cost benefits [9]

Core Viewpoint - Kuaishou has fully launched its AI model-driven recommendation system, OneRec, which is the first industrial-grade recommendation solution in the industry, setting a new standard globally [1][15]. Group 1: Technological Advancements - Kuaishou's technology has reached a top-tier level, particularly in video generation models [2]. - The company has made significant underlying technological advancements that surpass initial perceptions of it being merely a short video platform [3]. Group 2: Recommendation System Overview - Recommendation systems are a major technological innovation of the mobile internet era, utilized by popular platforms like Kuaishou, Douyin, and Pinduoduo [4]. - Traditional recommendation systems typically rely on user-based collaborative filtering and content-based collaborative filtering [4][6]. Group 3: Challenges in Traditional Systems - Traditional multi-stage recommendation systems face issues such as low overall GPU utilization and inefficiencies due to independent model operations [10][11]. - The complexity of user interests and the conflicting goals of increasing click-through rates while maintaining content diversity lead to decreased recommendation accuracy [9][10]. Group 4: OneRec's Innovations - OneRec shifts from a multi-stage filtering approach to an end-to-end model that directly generates a list of recommended videos based on user interests [16]. - The system employs a multi-modal semantic tokenizer to deeply understand video content beyond surface-level tags, enhancing content comprehension [21][24]. Group 5: User Modeling and Interest Tracking - OneRec integrates user behavior over time to create a comprehensive "interest sequence," allowing for more accurate recommendations that adapt to changing user preferences [28][30]. - The model uses deep neural networks to automatically learn complex interest changes from large datasets, improving recommendation accuracy [30]. Group 6: Recommendation Generation - The system utilizes an encoder-decoder structure, where the encoder compresses user interest trajectories into vectors, and the decoder generates a sequence of recommended content [32][33]. - The introduction of a Mixture of Experts (MoE) architecture enhances model capacity and efficiency, allowing for personalized recommendations while maintaining content diversity [34][36]. Group 7: Reinforcement Learning Integration - OneRec incorporates a reward mechanism using reinforcement learning to align user preferences with recommendation outcomes, enhancing the overall effectiveness of the system [38][44]. - The model's training includes various reward signals to ensure a balanced distribution of content types and to adapt to real-world business complexities [41][42]. Group 8: Performance Metrics - During the testing phase, OneRec demonstrated performance metrics comparable to existing complex systems, with user engagement metrics such as watch time and user lifecycle showing positive growth [46][47]. - In local life scenarios, OneRec achieved a 21% increase in GMV and significant growth in order volume and new customer acquisition [48]. Group 9: Future Considerations - Despite its advancements, OneRec still faces challenges related to inference speed, resource consumption, and further optimization of the reward mechanism [49]. - The introduction of OneRec marks a new phase in recommendation systems, aligning them with the latest advancements in AI and machine learning [49][50].

Core Viewpoint - The article discusses the transformation of recommendation systems through the integration of large language models (LLMs), highlighting the introduction of the "OneRec" system by Kuaishou, which aims to enhance efficiency and effectiveness in recommendation processes [2][35]. Group 1: Challenges in Traditional Recommendation Systems - Traditional recommendation systems face significant challenges, including low computational efficiency, conflicting optimization objectives, and an inability to leverage the latest AI advancements [5]. - For instance, Kuaishou's SIM model shows a Model FLOPs Utilization (MFU) of only 4.6%/11.2%, which is significantly lower than LLMs that achieve 40%-50% [5][28]. Group 2: Introduction of OneRec - OneRec is an end-to-end generative recommendation system that utilizes an Encoder-Decoder architecture to model user behavior and enhance recommendation accuracy [6][11]. - The system has demonstrated a tenfold increase in effective computational capacity and improved MFU to 23.7%/28.8%, significantly reducing operational costs to just 10.6% of traditional methods [8][31]. Group 3: Performance Improvements - OneRec has shown substantial performance improvements in user engagement metrics, achieving a 0.54%/1.24% increase in app usage duration and a 0.05%/0.08% growth in the 7-day user lifecycle (LT7) [33]. - In local life service scenarios, OneRec has driven a 21.01% increase in GMV and an 18.58% rise in the number of purchasing users [34]. Group 4: Technical Innovations - The system employs a multi-modal fusion approach, integrating various data types such as video titles, tags, and user behavior to enhance recommendation quality [14]. - OneRec's architecture allows for significant computational optimizations, including a 92% reduction in the number of key operators, which enhances overall efficiency [27][28]. Group 5: Future Directions - Kuaishou's technical team identifies areas for further improvement, including enhancing inference capabilities, developing a more integrated multi-modal architecture, and refining the reward system to better align with user preferences [38].