[Table_IndNameRptType] 全球科技 行业周报 DeepSeek 线上模型版本升级至 V3.1，快手发布业绩派发股息 [Table_IndRank] 行业评级：增持 报告日期： 2025-08-25 [Table_Chart] 行业指数与沪深 300 走势比较 执业证书号：S0010521080002 邮箱：jinrong@hazq.com 分析师：刘京松 执业证书号：S0010524090001 邮箱：liujs@hazq.com 分析师：赵亮 执业证书号：S0010525070001 邮箱： zhaoliang@hazq.com 相关报告 1.H20 恢复在华销售，监管约谈外卖平 台 2025-07-21 2.AI 系列专题跟踪——视频及图像生 成模型 2025-07-15 3.模型与应用共振，关注 AI 应用 2025 -07-13 主要观点： [⚫Table_Summary] 本周行情回顾 分析师：金荣 2.国内 AI： ⚫ DeepSeek 线上模型版本升级至 V3.1 8 月 19 日，DeepSeek 官方发布通知：线上模型版本已升级 至 V3.1，上下文长度拓展至 128k， ...

用300多元的成本，就能制作一部科幻短片？过去需要一整支团队才能完成的视频制作，如今只需要一张图、一句话，就能生成电影级质感的视频。AI 视频生成技术正以惊人的速度走向规模化应用，引发互联网巨头、初创企业的激烈竞逐。 "密码0821，快去重启，消灭他们。""收到！"当男主角接到指令，历尽艰难输入重启控制台的密码，在暗无天日如同世界末日般的场景下，给人类带来 巨大威胁的异形生物在瞬间崩成了碎粒、消失不见……8月21日，曾参与《2012》《黑客帝国3》《变形金刚3》等多部好莱坞大片及国产科幻剧《三 体》视效工作的国际知名视效指导姚骐展示了一段自己用AI制作的2分钟科幻短片《归途》。 数百元做出影视级短片 姚骐创作《归途》所用的AI工具是来自北京的视频大模型。当天，百度"蒸汽机"（MuseSteamer）音视频一体化模型在发布不到50天后宣布重大升级。 与业界大量AI生成视频大多为"默片"不同，"蒸汽机"在行业内首次实现多人有声音视频一体化生成。百度商业体系商业研发总经理刘林告诉记者，使用 百度"蒸汽机"音视频一体模型制作的《归途》，整体成本仅330.6元人民币。而姚骐表示，如果这部短片是一部纯实拍的片子，成本可能 ...

Key Points - The article discusses various developments in the tech and automotive industries, highlighting significant corporate actions, product launches, and market strategies. Group 1: Corporate Developments - Former CFO of Zhibo Network publicly criticized the company's upcoming IPO, stating that he left due to a lack of confidence in the business and accused certain executives of being opportunistic [4][6]. - Alibaba announced the spin-off of Zhibo Network for an independent listing on the Hong Kong Stock Exchange, with plans to retain over 30% ownership post-IPO [6]. - Alibaba's Lingxi Entertainment has shifted its reporting structure to report directly to CFO Xu Hong, indicating potential changes in business strategy [12][13]. Group 2: Product Launches and Innovations - NIO unveiled the new ES8 model, with a starting pre-sale price of 416,800 yuan, featuring significant upgrades in size and technology [19]. - Vivo introduced the Vision Exploration Edition, the lightest MR headset in the industry, weighing only 398g, designed for enhanced user experience [30]. - DeepSeek released version 3.1, which includes significant upgrades and price adjustments for its API services, reflecting a shift towards next-generation domestic chips [11]. Group 3: Market Strategies - Alibaba's local services division is launching a new group-buying feature called "Flash Group," aimed at price-sensitive consumers, to compete with Meituan's similar offerings [18]. - Multiple ride-hailing platforms, including Didi and T3, have announced reductions in commission rates to support driver income and expand platform capacity [24][25]. - Zero Run Auto reported a cumulative delivery of over 900,000 vehicles, achieving profitability in the first half of the year and adjusting its annual sales target upwards [26][27]. Group 4: Financial Performance - Kuaishou reported a revenue of 35.05 billion yuan for Q2 2025, with a net profit increase of 20.1%, and announced a special dividend for shareholders [39]. - Bilibili's Q2 revenue reached 7.34 billion yuan, with significant growth in advertising and gaming revenue, and a record high in user engagement metrics [40]. Group 5: Competitive Landscape - Samsung's HBM4 samples have passed initial testing with Nvidia and are set to enter pre-production, potentially challenging SK Hynix's dominance in the AI memory chip market [44][45]. - Intel is negotiating with large investors to replicate a previous financing deal with SoftBank, aiming to bolster its capital structure [46]. Group 6: Privacy and Regulatory Issues - Meta is facing allegations of circumventing Apple's privacy restrictions to enhance ad revenue, with claims of misleading advertisers about the performance of its Shop Ads [51][52]. - xAI's Grok platform experienced a significant privacy breach, exposing over 370,000 user chat records due to design flaws in its sharing functionality [46][47].

参与过电影《2012》、《黑客帝国3》等好莱坞大片的视效指导姚骐今天公布了他用AI制作的科幻短片《归途》。 短片里，如同末日的世界里，巨大的异形生物追击驾驶汽车的人类、巨型蜘蛛爬在高楼等场景栩栩如生。姚骐评价"（效果）跟实拍差不 多。" 他向第一财经等记者透露，整部短片用了40多个镜头，每个镜头生成3次，共计120个视频片段，其中包括18个10秒一体化的有声片段和102个 五秒片段，最终花费约一周时间制作完成。 姚骐说，如果这部短片是一部纯实拍或者CG制作的片子，可能需要几百万的成本。在好莱坞做镜头，有些复杂的镜头仅一个就要几十万甚至 上百万。此外，实拍还受限于场景实现难度、危险性以及演员、剧组成本，而AI技术的介入为创意实现提供了全新可能。 几百万实拍成本的短片，如果用AI生成，花费是多少？ 姚骐AI短片的合作对象、百度商业体系商业研发总经理刘林告诉记者，该片使用百度蒸汽机音视频一体模型，整体成本约在330.6元人民币。 但AI生成还有不少进步空间 目前，百度视频生成模型上线50天，最大的用户来自百度内部，包括搜索业务、移动生态创作者等，其次是专业领域创作者，以及企业客 户。 眼下视频生成赛道已足够卷。快手 ...

Core Viewpoint - The global large model industry continues to develop rapidly post "DeepSeek innovation heat," with an acceleration in model iterations and an increase in computing power demand driven by token consumption [2][7][25]. Group 1: Global Model Updates and Computing Demand - In Q2 2025, major model companies like Google and OpenAI released significant updates, including OpenAI's GPT-5, which improved efficiency and reduced API costs, thus increasing computing power demand [3][13][25]. - The release of GPT-5 marked a shift towards efficiency, with a notable reduction in token consumption and a context window expansion to 400K tokens, enhancing application capabilities and driving further demand for computing resources [18][22][25]. - The North American model updates have created a preliminary closed loop in computing power demand, with companies like Google and Anthropic seeing rapid increases in token consumption [3][30]. Group 2: Domestic Model Development and Market Dynamics - Domestic companies, while still trailing behind in model capabilities, have made significant strides since 2025, with firms like ByteDance and Kimi releasing updated models that have increased computing power consumption [4][36]. - The domestic AI chip industry is evolving from single-chip solutions to system-level designs, supporting the iteration and deployment of large models [4][43]. - The anticipated updates from open-source models like DeepSeek in Q3 2025 could reignite investment sentiment in the domestic AI industry [4][36]. Group 3: Token Consumption Trends - Token consumption has surged globally, with major players like Google, Microsoft, and ByteDance experiencing significant increases in token usage since 2025 [27][30]. - Google's AI Overview feature has been a key driver of its token consumption growth, leveraging its vast user base to generate high-frequency AI summaries [30][31]. - The current market dynamics reflect a balance between free access to AI technologies and the monetization of high-value applications, with paid products showing a clear differentiation in performance and reliability [34][35]. Group 4: Future Outlook and Investment Opportunities - The ongoing model updates and the increasing efficiency of token usage are expected to drive sustained growth in computing power demand, with both cloud and edge computing becoming critical [22][24][36]. - The competitive landscape suggests that companies with robust financial backing, like Google and Meta, will continue to push for model updates, further enhancing computing demand [26][30]. - The domestic AI industry is poised for growth as local firms enhance their capabilities and seek to capture market share in the evolving landscape of AI applications [4][43].

Investment Rating - The report maintains a "Buy" rating for several companies in the media sector, including Kuaishou, Giant Network, and Perfect World [5][8]. Core Insights - The report highlights a 14.08% year-on-year growth in the gaming market for the first half of 2025, with a market size of 168 billion yuan and a user base of approximately 679 million [2][26]. - The AI application sector is experiencing significant developments, with Tencent's mixed reality model and various new AI features being launched [2][42]. - The film industry saw a total box office of 1.541 billion yuan in the last week, with "Nanjing Photo Studio" leading the box office [39][40]. Market Performance - The media industry (Shenwan) rose by 1.13% from July 26 to August 1, outperforming the Shanghai Composite Index, which fell by 0.94% [11][17]. - Among the media sub-sectors, the film industry saw the highest growth at 3.22%, while digital media experienced a slight decline of 0.06% [11][17]. Key Data and Dynamics AI Applications - Estimated iOS downloads for AI applications during the week of July 26 to August 1 were 412,700 for Deepseek, 2,077,300 for Doubao, 725,000 for Quark, and 570,700 for Tencent Yuanbao, with varying week-on-week changes [21][25]. Gaming Sector - The mobile gaming market reached 125.3 billion yuan in the first half of 2025, growing by 16.55% year-on-year, with self-developed games generating over 11% growth in overseas revenue [26][28]. - Top games in the sales rankings as of July 31 included "Honor of Kings," "Peace Elite," and "Delta Action" [32][34]. Film Sector - The top three films for the week of July 25 to July 31 were "Nanjing Photo Studio," "Chang'an's Lychee," and "The Stage," with respective box office earnings of 808.4 million yuan, 211.3 million yuan, and 183.7 million yuan [39][40]. Investment Recommendations - The report recommends focusing on themes such as AI applications and cultural exports, with particular attention to gaming, IP, short dramas, and publishing sectors [3][44]. - Specific companies highlighted for investment include Kuaishou, Giant Network, and Shanghai Film [3][44].

Core Insights - AI-generated ASMR videos are creating a global trend on short video platforms, driven by the rapid growth of accounts specializing in this content, such as a fruit-cutting ASMR video that gained over 100,000 followers in just five days on TikTok [1][6] - The integration of AI technology with ASMR has led to unique sensory experiences, allowing for the creation of visually and audibly captivating content that defies physical limitations [6][14] - Major companies like Kuaishou, ByteDance, MiniMax, and Baidu are accelerating their investments in AI ASMR, leading to diverse monetization models ranging from creator tips to platform commercialization [6][31] Industry Trends - The emergence of AI ASMR videos has been categorized into various types, including unconventional fruit-cutting videos, immersive eating broadcasts, and therapeutic clay ASMR videos [7][9][10] - The Google Veo3 model has significantly advanced the field of AI video generation by enabling seamless audio-visual synchronization, thus lowering the barriers for content creation [14][20] - The competitive landscape includes various AI video generation tools, with companies like Kuaishou and ByteDance rapidly developing their own models to capture market share [20][38] Monetization Opportunities - AI ASMR creators are capitalizing on the low production costs and high engagement rates, with some earning substantial income through selling custom sound packs and video tutorials [31][33] - The rapid growth of followers and engagement on platforms like TikTok has led to significant advertising revenue opportunities for creators, with some achieving monthly incomes exceeding 10,000 yuan [31][33] - Companies are also exploring advertising integration within AI ASMR content, making it an attractive marketing avenue for brands in various sectors [33][38] Future Outlook - The AI video generation market is expected to reach a tipping point for commercialization by 2025, with significant revenue growth anticipated for leading products [37][39] - Kuaishou's Keling AI has already generated over 100 million yuan in revenue within nine months, indicating a strong potential for commercial success in the AI video generation space [35][38] - The ongoing technological advancements and evolving user preferences will determine which companies can maintain a competitive edge and achieve sustainable profitability in the AI ASMR market [39][40]

Key Insights - The report highlights the importance of geopolitical factors and AI as catalysts for investment opportunities in overseas markets [1][2] - The upcoming release of OpenAI's GPT-5 is expected to enhance AI capabilities across various applications, potentially impacting multiple sectors [3][10] - The report recommends a diversified portfolio of stocks, focusing on companies with strong growth potential in AI, consumer goods, and real estate [4][20] Recent Key Events - The third round of US-China trade talks took place in Sweden, with significant global attention on the outcomes [1][8] - The World Artificial Intelligence Conference (WAIC 2025) in Shanghai gathered over 1,572 leaders and scholars to discuss the future of AI [2][9] - The H20 chip export ban was lifted, allowing for renewed trade with China, which could influence tech companies [2][9] Market Situation - The Hong Kong and US stock markets saw significant gains in July, with the Hang Seng Index rising from 24,072 points to 25,524 points, a 6.0% increase [11] - Net inflows from southbound trading reached 110.8 billion HKD in July, indicating strong investor interest in Hong Kong stocks [12][13] Current Portfolio Recommendations - **Pop Mart (9992.HK)**: Strong growth in overseas business with a 475%-480% increase in international revenue in Q1 2025 [21][22] - **Jintai Holdings (2228.HK)**: Positioned as a leader in AI for Science, with significant growth potential in the pharmaceutical sector [26][27] - **China Qinfa (0866.HK)**: Improved balance sheet with a net profit increase of 150.5% in 2024, driven by loan restructuring [30][33] - **Greentown China (3900.HK)**: Despite a decline in profits, the company is expected to stabilize and lead the industry due to strong land acquisition strategies [36][39] - **Alibaba (9988.HK)**: Revenue growth of 7% in Q4 2025, with a strong focus on AI and cloud services [40][41] - **Kuaishou-W (1024.HK)**: Significant growth in e-commerce and advertising revenue, driven by innovative marketing strategies [44][46] - **Xiaomi Group-W (1810.HK)**: Record revenue of 111.3 billion CNY in Q1 2025, with a strong performance in both mobile and AIOT sectors [47]

Group 1 - GLM-4.5 is an open-source model designed for agents, excelling in reasoning, coding, and agent tasks, with leading performance in domestic tests [1] - The model employs a mixed expert architecture, offering two modes with high parameter efficiency, achieving performance comparable to larger competitors [1] - It features low cost (0.8 yuan per million tokens) and high speed (up to 100 tokens per second), supporting full-stack development tasks [1] Group 2 - Yuntian Lifa is focusing entirely on AI inference chips, aiming to enhance single-chip computing power to thousands of TOPS by 2028 to support trillion-parameter large models [2] - The company utilizes an innovative "computing power building block" architecture with fully domestic technology, compatible with mainstream open-source models and the HarmonyOS [2] - The strategy includes a triad layout of edge, cloud, and intelligent machines, forming four major business segments targeting edge computing, cloud-based large model inference, and intelligent machines [2] Group 3 - Coze has open-sourced two core products (Coze Studio and Coze Loop) under the Apache 2.0 license, receiving 9.5K stars on GitHub [3] - Coze Studio offers a no-code development platform allowing users to create agents through drag-and-drop operations, supporting multi-platform deployment; Coze Loop provides a full lifecycle management toolchain [3] - The open-source strategy aims to establish a new paradigm for agent development, providing a complete toolchain and flexible customization capabilities [3] Group 4 - Kuaishou's Keling AI has released significant updates, including a "spiritual canvas" supporting five-person collaborative creation and a greatly enhanced "multi-image reference" feature [4][5] - The new multi-image reference function addresses consistency issues in AI video generation, showing a 102% improvement in blind tests regarding character representation, dynamic quality, and artistic style stability [5] - A new local reference feature allows users to precisely define reference areas, making video generation results more controllable and significantly lowering the barrier for daily creative video production [5] Group 5 - Lovart, the world's first design agent, has officially launched, utilizing Tencent's Mix Yuan 3D model API for ultra-high-definition detail modeling [6] - The Mix Yuan 3D v2.5 version employs a sparse 3D native architecture, achieving a tenfold increase in geometric model accuracy compared to previous generations, supporting 4K PBR texture mapping [6] - The Mix Yuan strategy remains open-source, with plans for multiple upgrades by 2025, and has surpassed 2.3 million downloads on the Hugging Face platform, having also open-sourced the Mix Yuan 3D World Model 1.0 [6] Group 6 - Alibaba has open-sourced the Tongyi Wanshang Wan2.2 video generation model, the first in the industry to use the MoE architecture, with a total of 27 billion parameters, saving 50% in computing resources [7] - The new model introduces a cinematic aesthetic control system, offering over 60 parameters to adjust lighting, composition, and color [7] - The 5 billion version of the unified video generation model supports both text-to-video and image-to-video generation, deployable on consumer-grade graphics cards [7] Group 7 - SenseTime has launched the Wuneng Embodied Intelligence Platform, providing robots with perception, navigation, and multimodal interaction capabilities based on world models, addressing data bottlenecks [8] - The Wuneng platform can generate high-quality simulation data that adheres to physical rules and offers first and third-person perspectives, enhancing robot training efficiency [8] - This platform empowers robots with intelligent interaction capabilities, demonstrated by a robot that can present PowerPoint slides, showcasing global memory capabilities and transitioning from a tool to a partner in interaction [8] Group 8 - The Shanghai Institute of Science Intelligence, Fudan University, and Infinite Light Year have jointly launched the "Galaxy Enlightenment Scientific Intelligence Open Platform," providing AI-enabled full-link research tools for scientists [10] - The platform is designed with a "scientist-centered" approach, integrating over 200 scientific models across 12 disciplines and 12PB of high-value scientific data, attracting over 120 research teams [10] - It offers six core capabilities: native intelligent agent scientific exploration engine, universal scientific model repository, efficient scientific computing, wet and dry experiment closed-loop, high-value scientific data, and a multidisciplinary collaborative research community, marking the entry into the 2.0 era of scientific intelligence [10] Group 9 - Shopify announced its "All in AI" strategy, sharing successful implementation experiences three months post-announcement, emphasizing universal AI usage without cost limits and default legal team support [11] - The company has built a unified AI entry point, connecting all internal tools via an MCP server, allowing employees to freely construct workflows, significantly enhancing departmental efficiency [11] - Shopify employs a counterintuitive strategy by encouraging AI to demonstrate its thought process rather than hiding it, hiring more junior talent as "AI natives," increasing prototype creation, and linking AI usage to employee performance [11] Group 10 - OpenAI's board chair Bret Taylor believes the SaaS applications of 2010 will evolve into intelligent agent companies by 2030, indicating we are in an "accelerated internet bubble era" [12] - The AI market is divided into three main areas: frontier large models (high competition, difficult entry), AI tools (challenging but with opportunities), and application-layer AI (the greatest opportunity) [12] - Entrepreneurship requires a core "argument" rather than blindly "failing fast," with true customer value for B2B companies needing market validation, as the market explores the "LAMP" technology stack in the AI era, with future intelligent marginal costs approaching zero [12]

Core Viewpoint - During the WAIC, Kuaishou's generative AI launched significant upgrades to its Ling Animation Canvas and multi-image reference features, enhancing creative capabilities for users [1] Group 1 - The Ling Animation Canvas serves as a creative workspace, integrating unlimited visualization space, intelligent creative assistance, and real-time collaboration for creators [1]