策略研究 | 证券研究报告 — 点评报告 2025 年 7 月 13 日 策略点评 国产算力产业链贯通，行情催化在即 国产算力产业链"产业突破-业绩验证-需求验证"逻辑链条已逐步打通，产业 链行情正在迎来催化。 tianran.gao@bocichina.com 证券投资咨询业务证书编号：S1300522100001 ◼ 国产算力产业链"产业突破-业绩验证-需求验证"逻辑链条已逐步打通，产 业端，沐曦集成、摩尔线程两大国产 GPU 厂商科创板 IPO 获受理，填补 A 股全功能 GPU 空白，华为算力集群性能媲美英伟达，产业链重点公司 业绩来看，工业富联中报业绩创历史同期新高，需求端，国产大模型 Tokens 消耗量高速增长。国产 GPU 上市潮、硬件厂商业绩兑现与 AI 大 模型需求增长形成共振，国产算力产业从技术研发、商业落地到需求支撑 已进入高速增长周期，全产业链自主化进程提速，国产算力行情亦有望受 到催化，建议关注国产算力芯片、服务器、PCB、光通信厂商等。 ◼ 风险提示：1）政策落地不及预期，宏观经济波动超预期。2）市场波动风 险。3）海外经济超预期衰退、流动性风险。 策略点评 国产算力产业链"产业 ...

Core Insights - ByteDance is focusing on "emotional large models" to provide API calls and AI dialogue solutions for enterprises, indicating a strategic shift towards enhancing user emotional experiences in AI interactions [1][2][4] - The development of "emotional large models" is seen as a significant trend in AI, moving from mere tools to emotional companions, which opens new application scenarios [5][7] Group 1: Emotional Large Models Overview - "Emotional large models" differ from traditional chatbots by emphasizing emotional understanding and user experience, utilizing voice tone, pauses, and expressions to generate appropriate responses [3][4] - The technology evolution of "emotional large models" is driven by two paths: enhancing multimodal emotional computing capabilities on general large models and focusing on generative models specifically for emotional applications [5][6] Group 2: Market Trends and Growth Potential - The AI companionship market is expected to see explosive growth, with the number of active users increasing 30 times from 2018 to 2023, and the global market size projected to rise from $30 million to $150 billion between 2023 and 2030, with a CAGR of 236% [7] - Character.AI exemplifies the potential of "emotional large models" by enabling interactive AI character experiences, with significant user engagement reflected in its mobile downloads and web traffic [8][10] Group 3: Technical Aspects and Challenges - "Emotional large models" require more NLP experts and have different parameter and computational needs compared to traditional models, with training requiring 30%-50% more computational power [10][11] - The current gap in development between domestic and international "emotional large models" indicates that domestic advancements are approximately one year behind [11] Group 4: ByteDance's Strategic Positioning - ByteDance plans to leverage various vertical large models to double the monthly active users of its product Doubao by 2025, focusing on entertainment, social, and gaming scenarios [14] - The integration of "emotional large models" with hardware like smart speakers and AI companions is part of ByteDance's strategy to enhance user interaction and experience [14][15]

Core Insights - The article highlights the launch of a new financial model solution by Huoshan Engine, focusing on enhancing customer service and internal efficiency for financial institutions through AI-native applications and digital employees [1] - The financial sector is identified as a leading area for the large-scale application of AI models, with Huoshan Engine's collaboration with financial institutions accelerating since the release of the Doubao model in May 2024 [1] - Huoshan Engine's AI solutions have already reached 70% of systemically important banks and numerous brokerage and fund companies, emphasizing the importance of customer service and internal efficiency in the financial industry [1] Financial AI App - The AI App serves as a mobile platform that utilizes AI models to provide intelligent investment advice, trading, services, and information interactions, already implemented in several brokerage firms [3] - The app aims to transform the user experience of financial applications [3] Capability Framework of AI App - The AI App solution is structured into three layers: - The intelligent interaction layer includes financial clients and Doubao App, featuring expert intelligent agents like market assistants [4] - The middle platform capability layer integrates intelligent control, data-knowledge-intelligence analysis, and model evaluation to ensure the iterative development of AI applications [4] - The infrastructure layer is built on the Doubao model, AI cloud-native security, and computing power management [4] - The Huoshan Engine Data Agent has effectively created investment advisory intelligent agents for multiple enterprises, enabling users to access multi-dimensional market analysis and real-time tracking of trends and fund movements [4] Digital Employees - The digital employee solution encompasses six key scenarios, creating an inclusive AI tool matrix: - Business assistants automate the entire due diligence cycle, shifting from experience reliance to AI-driven processes [5] Intelligent Customer Service and Risk Control - Intelligent customer service integrates ticket extraction, AI Q&A, and quality inspection to enhance service efficiency and compliance [6] - Intelligent risk control replaces traditional OCR with VLM to automatically extract multi-modal risk factors, upgrading risk insight models [6] Collaborative Ecosystem - The article discusses the importance of ecosystem collaboration for the large-scale implementation of AI-native applications, with Huoshan Engine signing partnerships with ten financial technology pioneers to co-create a new financial intelligence ecosystem [8] - The collaboration aims to drive innovation and accelerate the financial industry's transition towards intelligence [8]

国内有众多 AI 大模型研发机构，如百度、阿里、字节跳动、科大讯飞等，从这些机构的实践来看，大模型主要有以下发展方向： ·行业深度赋能：科大讯飞计划将星火大模型从 "通用" 走向 "行业"，深度赋能汽车、教育、医疗、智慧城市、赛事运营等产业。百度、阿里等公司的大模型 也在金融、工业、政府、科研、电商等领域积极探索应用，未来大模型会针对不同行业的特点和需求，进行定制化开发和优化，为各行业提供更专业、更精 准的服务，推动行业智能化升级。 ·智能应用创新：随着大模型技术的发展，将催生更多新型智能应用。字节跳动提出 "互联网正从 APP 时代进入 Agents 时代"，其火山引擎发布的方舟平台 等系列工具，构建了服务 Agent 开发的完整体系，展现出 AI 重构软件开发范式的潜力，也催生了不少如他她它、推氪AI等的应用层面的产品。未来，大模 型将与更多新兴技术结合，创造出如智能助手、智能创作工具等更多创新应用，改变人们的生活和工作方式。 生态建设方向 ·开源共享：开源成为大模型发展的重要趋势。2025 年以来，字节豆包、百度文心、阿里通义千问等均推出开源模型。商汤科技发布了 LazyLLM 开源框架 等产品，Min ...

7月1日，根据 IDC的最新报告，2024年全年阿里云AI IaaS市场份额达23%排名第一，超过第二名和第三名总和。 华为云占比10%排名第二，火山引擎占比9%排名第三，中国电信占比8%排名第四，腾讯云、百度云和aws中国、 商汤皆占比7%。这也是IDC首次以AI IaaS进行数据统计。 在今年3月的2024年第四季度及全年业绩年报媒体沟通会上，腾讯高管提到，2024年腾讯年度资本支出达107亿美 元，相当于同期收入的大约12%。其中第四季度资本支出增加十分显著，因为腾讯在这一季度购买了很多GPU。 2025年腾讯计划进一步增加AI相关资本支出，预期资本支出占收入的比例为低两位数百分比。 采写：南都记者 林文琪 聚焦到生成式AI IaaS（GenAI IaaS）市场中，报告显示阿里云取得模型训练和模型推理市场的双项冠军。另外， 2024年模型训练消耗的生成式AI IaaS资源约为模型推理的3.25倍，随着"爆款"大模型应用的出现，IDC预计AI算力 有望在2025年进入"训推"拐点，推动AI算力需求从训练驱动过渡到推理驱动。IDC认为，推理场景的爆发将为国 产算力未来发展带来更多市场机会，互联网、运营商、自 ...

Core Viewpoint - The article discusses the transformation of AI infrastructure, emphasizing the need for a heterogeneous computing architecture that integrates both CPU and GPU resources to meet the demands of large AI models and their applications [2][4][7]. Group 1: AI Infrastructure Transformation - AI large models are reshaping the computing landscape, requiring organizations to rethink their AI infrastructure beyond just adding more GPUs [2]. - The value of CPUs, long underestimated, is returning as they play a crucial role alongside GPUs in AI workloads [3][4]. - A complete AI business architecture necessitates the simultaneous upgrade of both CPU and GPU resources to fulfill end-to-end AI business needs [5][7]. Group 2: Challenges and Solutions - The rapid iteration of large language models presents four main challenges for processors: low GPU computing efficiency, low CPU utilization, increased data movement bandwidth requirements, and GPU memory capacity limitations [5]. - Intel has developed various heterogeneous solutions to address these challenges, including: - Utilizing CPUs in the training and inference pipeline to reduce GPU dependency, improving overall training cost-effectiveness by approximately 10% [6]. - Optimizing lightweight models with the Xeon 6 processor to enhance responsiveness and free up GPU resources for primary models [6]. - Implementing QAT hardware acceleration for KV Cache compression, significantly reducing loading delays and improving user response times [6]. - Employing a sparse-aware MoE CPU offloading strategy to alleviate memory bottlenecks, resulting in a 2.45 times increase in overall throughput [7]. Group 3: Intel's Xeon 6 Processor - Intel's Xeon 6 processor, launched in 2024, represents a comprehensive solution to the evolving demands of data centers, featuring a modular design that decouples I/O and compute modules [9][10]. - The Xeon 6 processor achieves significant performance improvements, with up to 288 physical cores and a 2.3 times increase in overall memory bandwidth compared to the previous generation [12]. - It supports advanced I/O capabilities, including a 1.2 times increase in PCIe bandwidth and the first support for CXL 2.0 protocol, enhancing memory expansion and sharing [13]. Group 4: Cloud and Local Deployment Strategies - The trend of enterprises seeking "local controllable, performance usable, and cost acceptable" AI platforms is emerging, particularly in sectors like finance and healthcare [24]. - Intel's high-cost performance integrated machine aims to bridge the gap for local deployment of large models, offering flexible architectures for businesses [25][26]. - The integrated machine solution includes monitoring systems and software frameworks that facilitate seamless migration of existing models to Intel's platform, ensuring cost-effectiveness and maintainability [28][29]. Group 5: Collaborative AI Ecosystem - The collaboration between Intel and ecosystem partners is crucial for redefining the production, scheduling, and utilization of computing power, promoting a "chip-cloud collaboration" model [17][30]. - The introduction of the fourth-generation ECS instances by Volcano Engine, powered by Intel's Xeon 6 processors, showcases the enhanced performance capabilities in various computing scenarios [18][20].

Group 1 - In 2025, global enterprises are undergoing a "leadership test" driven by AI, with companies like Microsoft, Huawei, and ByteDance leading the way in productivity and revenue growth [1][2] - Traditional leadership structures are being challenged as AI becomes integrated into the workplace, leading to a shift towards intelligent organizations [2][3] - A significant percentage of young leaders (82%) are already using AI in their work, indicating a generational shift in leadership dynamics [3][7] Group 2 - The reliance on AI tools is creating a power shift within organizations, as employees prefer consulting AI over their leaders due to its impartiality and efficiency [7][8] - The leadership environment is changing, with younger employees using AI to challenge traditional decision-making processes, leading to a potential "digital divide" within teams [10][11] - Decision-making transparency is increasing, putting pressure on leaders as AI can analyze and critique their decisions [11][12] Group 3 - Leaders must redefine their roles to leverage AI effectively, focusing on collaboration between human and machine [13][14] - Trust in leadership is being tested as employees increasingly rely on AI for decision-making, raising questions about the rationality of human leaders [14][16] - The concept of "symbiotic leadership" is emerging, where leaders must integrate traditional and digital economies, creating value through collaboration [17][18] Group 4 - Companies are encouraged to adopt a leadership operating system that addresses six key pain points, enhancing decision-making and team communication through AI [21][22] - Successful organizations are those that utilize platforms to integrate resources and foster collaboration within their ecosystems [18][19] - The future leader will be a coordinator of human-machine collaboration, focusing on understanding both data and human nature [29][30]

Core Insights - The article highlights the evolution of user demand for AI in smart terminals, shifting from "novelty experience" to "deep needs," emphasizing the desire for both engaging and efficient AI services [1] - Volcano Engine maintains a dominant position in the Chinese public cloud large model market with a 46.4% market share as of 2024, showcasing significant growth in AI capabilities [1][3] - The collaboration between Volcano Engine and smart terminal manufacturers aims to enhance user AI experiences by addressing pain points and optimizing large model capabilities [3][4] Market Position and Performance - Volcano Engine's daily token usage for its Doubao large model exceeded 16.4 trillion by May 2023, marking a 137-fold increase since its launch [1] - The company is positioned for continued growth in 2025, reflecting its deep integration with various industries [3] Technological Advancements - The AI capabilities of smart terminals are evolving, with Volcano Engine's models supporting diverse applications across text, voice, and image domains [4] - AI assistants are transitioning from basic tools to intelligent partners, enhancing user interaction through advanced functionalities [4][9] Application Innovations - Honor's AI assistant "YOYO" utilizes Doubao large model for intelligent photo editing, allowing users to issue voice commands for personalized tasks [4] - Vivo's AI assistant "Blue Heart Little V" integrates Doubao large model for real-time internet information retrieval, improving natural interaction [6] - OPPO's AI intent search feature, developed with Volcano Engine, enhances search capabilities across various applications, refining user query responses [7] - Nubia's multi-modal assistant "Little Star" expands the boundaries of mobile AI interaction by creating a comprehensive interaction system [9] Future Outlook - Volcano Engine aims to continue exploring new models and technologies in collaboration with industry partners, facilitating the implementation of more AI application scenarios in domestic smart terminals [9]

提起字节跳动，你或许会先想到抖音、今日头条这些现象级的产品，但你可能不知道，在 AI 领域，这家公司 又一次 悄然完成了一场惊人的逆袭。 2025 年 6 月 11 日，在北京举行的火山引擎 Force 原动力大会上，字节跳动 CEO 梁汝波自信地宣布了他们在 AI 领域的最新成绩：旗下的豆包大模型 1.6 以"史上最低价格 + 顶尖推理能力" 杀入 企业级市场 ，火山引擎在中国公有云大模型调用量的市场份额更是高达 46.4% ，超过了百度与阿里之和， 稳坐行业头把交椅。 短短不到一年，这家昔日 AI 领域的"边缘玩家"凭借激进策略和快速迭代，成功逆袭阿里、百度等老牌云厂商，以接近二分之一的市场份额登顶国内大模型 服务市场。火山引擎总裁谭待现场一组数据更引发惊呼： 截至今年 5 月底，豆包模型日均调用已飙升至 16.4 万亿 tokens ，是发布初期的 137 倍！ 在 全球 AI应用5月产品榜上，豆包排名第四，超越了DeepSeek，同时在IOS下载排行榜排行第二 。 这一系列亮眼数字背后，字节跳动在 AI 领域究竟有哪些战略与实践？ 混沌 AI 君将 从商业、产品、组织 三 个视角，对字节跳动的 AI版 ...

Core Insights - The article highlights the rapid integration of AI technology into daily life, emphasizing the shift in user demand from "usable" to "practical" and "high-quality" applications of AI [1] - Companies are accelerating the incorporation of AI capabilities into their product offerings, expanding the boundaries of application scenarios to enhance user experience [1] Group 1: AI Technology Integration - The "Doubao" large model has become a common choice across various industries, with its daily token usage exceeding 16.4 trillion by May 2025, marking a 137-fold increase since its launch [1] - The collaboration between Huoshan Engine and smart terminal manufacturers is focused on expanding creative boundaries in AI creation, providing users with practical AI experiences [1] Group 2: Image Creation and Processing - The demand for high-quality photos is increasing as AI technology transforms the photography field, impacting the entire process from image creation to post-processing [2] - Huoshan Engine leverages the capabilities of the Doubao family of large models to enhance AI creative abilities in smart terminals, allowing non-professional users to achieve professional-level post-processing [2] Group 3: Product Innovations - Nubia's Z70S Ultra photographer version integrates the Doubao large model for a comprehensive AI creative service, enabling users to generate high-quality images and perform real-time style transformations [3] - Samsung's Galaxy S25 series features the "Drawing Assistant" app, which utilizes the Doubao large model to provide various image processing capabilities, creating a mobile AI image workstation [3] Group 4: Experience Enhancement - Vivo collaborates with Huoshan Engine to upgrade AI functionalities in its phones, introducing "vivo AI Time Capsule" for personalized video creation through AI-generated music and lyrics [5] - Honor builds a personalized service system using AI technology, offering features like style creation and intelligent editing to enhance user experience in image processing [6] Group 5: Market Position and Future Outlook - Huoshan Engine is projected to maintain the leading position in China's public cloud large model service usage, with a market share of 46.4% by 2024 [6] - The ongoing collaboration between Huoshan Engine and smart terminal manufacturers aims to deepen the application of large models and expand capabilities, enhancing user interaction experiences [6]