2025 年 10 月 10 日 陈伟光 S0340520060001 电话：0769-22119430 邮箱： chenweiguang@dgzq.com.cn 罗炜斌 S0340521020001 电话：0769-22110619 邮箱： luoweibin@dgzq.com.cn 超配（维持） 电子行业双周报（2025/09/26-2025/10/09） 行 业 OpenAI 动作频频，AI Infra 需求有望高增 投资要点： 资料来源：iFind，东莞证券研究所 本报告的风险等级为中高风险。 本报告的信息均来自已公开信息，关于信息的准确性与完整性，建议投资者谨慎判断，据此入市，风险自担。 请务必阅读末页声明。 电子行业 SAC 执业证书编号： 行情回顾及估值：申万电子板块近2周（09/26-10/09）累计上涨2.13%， 跑输沪深300指数0.39个百分点，在申万行业中排名第10名；板块10月 累计上涨2.18%，跑赢沪深300指数0.70个百分点，在申万行业中排名第 5名；板块今年累计上涨56.86%，跑赢沪深300指数37.17个百分点，在 申万行业中排名第3名。估值方面，截至10月9日，SW ...

点评： Sora2能力显著提升，APP登顶下载榜首位。北京时间10月1日，OpenAI正式推出Sora2视频生成模型， Sora2对物理世界理解能力进一步提升，能够执行跨越多个镜头的复杂指令，并支持音视频同步生成， 使角色语音、环境声与画面动作自然对应，整体沉浸感显著增强，被官方称为GPT3.5时刻。OpenAI同 时推出基于Sora2模型的iOS社交应用Sora，具有两大特色功能：1)Cameo：用户通过上传自己的音视频 片段进行验证后，可将自己的虚拟形象插入至AI生成的场景中，并且该形象可与朋友分享；2)Remix： 基于现有的视频，用户可以输入新的提示词进行二次创作。目前Sora APP仅面向美国、加拿大地区，上 线第四天即登顶苹果美国应用商店免费应用榜第一名。 OpenAI加大算力储备力度，AIInfra需求有望高增。OpenAI在10月7日开发者大会上宣布ChatGPT周活跃 用户已经达到8亿人，开发者数量已经达到400万名，API每分钟处理约60亿tokens(数据来源：机器之心 公众号)，同时宣布ChatGPT升级为"超级APP"，用户能够在对话中直接调用第三方应用。AI模型训练及 广泛应用将进 ...

计算机周报 20251007 Sora 2 发布利好的三大方向 2025 年 10 月 07 日 ⚫ 市场回顾 上周（9.29-9.30）沪深 300 指数上涨 1.99%，中小板指数上涨 2.48%，创业板 指数上涨 2.75%，计算机（中信）板块上涨 2.68%。板块个股涨幅前五名分别为： 初灵信息、汇金股份、当虹科技、开普云、ST 易联众；跌幅前五名分别为：*ST 东通、中威电子、安博通、金现代、数据港。 ⚫ 行业要闻 ⚫ 公司动态 ➢ 概伦电子：9 月 29 日消息，公司董事会披露，拟通过"发行股份+现金" 方式收购成都锐成芯微科技股份有限公司 100%股权及纳能微电子（成都）股份 有限公司 45.64%股权。本次交易总对价 21.7384 亿元，其中股份对价 11.84 亿 元、现金对价 9.89 亿元；向交易对方合计发行股份 67,774,455 股，发行价 17.48 元/股；交易对方涵盖向建军、海南芯晟等，所获股份锁定期为 12 个月或 36 个 月。业绩承诺方面，若交易于 2025 年实施，锐成芯微 IP 授权业务收入承诺分别 不低于 1.2096/1.4274/1.6843 亿元；同期纳能 ...

9月17日，腾讯全球数字生态大会Agent + Infra专场聚焦"智能体驱动云基础设施跃进升级"这一主题，汇 聚众多行业领军者，共同探索AI时代基础设施的技术突破与产业实践。腾讯云智算各产品负责人及行 业客户深入解读AI原生云架构、主动服务型基础设施、全链路安全体系等核心议题，同时发布多项重 磅成果，为智能体规模化落地筑牢技术底座。 在安全层面，腾讯云通过一整套大模型与智能体安全治理框架，为云上客户提供安全、可控、可靠、满 足监管要求的安全建设思路。方案贯穿模型选型、模型训练、推理部署、业务应用全流程，实现边界 API与用户输入/输出安全、模型运行环境安全防护及态势管理、智能体与MCP身份和特权防护、智能体 行为与意图安全管控、数据安全全流程安全等。 此外，李力还分享了从数字世界到现实世界的思考。机器人是AI影响现实世界的一个关键形态，但具 身智能与大语言模型的数据需求存在显著差异：一方面是具身的数据采集难度比较大，包括视觉、环 境、关节角状态等数据均要采集；另一方面，采集的数据与硬件高度绑定，进一步增加了数据采集的门 槛。因此，腾讯云希望能够帮助客户跳过数采、训练环节、直接提供支持跨本体直接使用的具身模型 ...

Core Insights - Ant Group's open-source team unveiled the 2.0 version of the "2025 Large Model Open Source Development Ecosystem Panorama" at the Shanghai Bund Conference, showcasing significant changes in the open-source landscape [2][4][10] Group 1: Ecosystem Changes - The updated panorama includes 114 projects, a decrease of 21 from the previous version, with 39 new projects and 60 projects that have exited the stage, including notable ones like TensorFlow, which has been overtaken by PyTorch [4][5] - The overall trend indicates a significant reshuffling within the ecosystem, with a median age of only 30 months for projects, highlighting a youthful and rapidly evolving environment [5][10] - Since the "GPT moment" in October 2022, 62% of the projects have emerged, indicating a dynamic influx of new entrants and exits [5][10] Group 2: Project Performance - The top ten most active open-source projects reflect a focus on AI, LLM, Agent, and Data, indicating the primary areas of interest within the ecosystem [7][9] - The classification framework has evolved from broad categories to more specific segments, including AI Agent, AI Infra, and AI Data, emphasizing the shift towards an "agent-centric" era [10][19] Group 3: Contributions by Region - Among 366,521 developers, the US and China contribute over 55%, with the US leading at 37.41% [10][12] - In specific areas, the US shows a significant advantage in AI Infra and AI Data, with contributions of 43.39% and 35.76% respectively, compared to China's 22.03% and 21.5% [12][14] Group 4: Methodological Evolution - The methodology for selecting projects has shifted from a known starting point to a broader approach that captures high-activity projects, increasing the threshold for inclusion [15][18] - The new methodology aligns with Ant Group's goal of providing insights for internal decision-making and guidance for the open-source community [15][18] Group 5: AI Agent Developments - The AI Agent category has evolved into a structured system with various specialized tools, indicating a transition from chaotic growth to systematic differentiation [19][21] - AI Coding has expanded its capabilities, covering the entire development lifecycle and supporting multimodal and context-aware functionalities [23][27] Group 6: Market Trends - The report predicts significant commercial potential in AI Coding, with new revenue models emerging from subscription services and value-added features [24][27] - Chatbot applications have seen a peak but are now stabilizing, with a shift towards integrating knowledge management for long-term productivity [28][30] Group 7: Infrastructure and Operations - The Model Serving segment remains a key battleground, with high-performance cloud inference solutions like vLLM and SGLang leading the way [42][45] - LLMOps is rapidly growing, focusing on the full lifecycle management of models, emphasizing stability and observability [50][52] Group 8: Data Ecosystem - The AI Data sector appears stable, with many projects originating from the AI 1.0 era, but is facing challenges in innovation and engagement [58][60] - The evolution of data infrastructure is anticipated, moving from static repositories to dynamic systems that provide real-time insights for models [60][61] Group 9: Open Source Dynamics - A trend towards customized open-source licenses is emerging, allowing for more control and flexibility in commercial negotiations [62][63] - The landscape of open-source projects is being challenged, with some projects operating under restrictive licenses, raising questions about the definition of "open source" [62][63] Group 10: Competitive Landscape - The competitive landscape is marked by a divergence between open-source and closed-source models, with Chinese projects flourishing while Western firms tighten their open-source strategies [67][68] - The introduction of MoE architectures and advancements in reasoning capabilities are becoming standard features in new models, indicating a shift in focus from scale to reasoning [69][70]

Core Viewpoint - The article highlights the advancements in domestic chip performance and the strategic focus of Tencent on enhancing AI computing power through a heterogeneous computing platform that integrates various chip resources [1][3]. Group 1: Chip Performance and Strategy - Tencent Cloud is leveraging a heterogeneous computing platform to provide AI computing power, which is fully compatible with mainstream domestic chips [1]. - The application ratio of domestic GPU chips is expected to increase by 2025, as cloud vendors maintain a diversified strategy for computing chips, including self-developed and domestic chip testing [3]. - Tencent's management emphasizes the importance of both imported and domestic chips, noting that some domestic chips can effectively handle smaller model inference tasks [3]. Group 2: Efficiency and Capital Expenditure - Tencent has made significant capital expenditures, totaling 83.1 billion yuan from Q4 of last year to Q2 of this year, to rapidly build computing power for self-developed products and client services [4]. - The company has invested heavily in optimizing computing efficiency, with improvements in various areas such as storage and communication, which can enhance operational efficiency significantly [4]. - The upgrade of AI infrastructure is expected to accelerate the large-scale deployment of intelligent agents, with rapid iterations in technology paradigms related to inference in the open-source community [4].

Core Viewpoint - The article discusses the current state and future potential of the AI industry, emphasizing the rapid technological changes and the uncertainty surrounding AI applications and entrepreneurship. It raises questions about whether early entrepreneurs can build a competitive edge or if they risk becoming obsolete due to fast-evolving technologies [2]. Group 1: AI Industry Development - The AI core industry in Haidian District is projected to exceed 280 billion yuan in 2024, with an annual growth rate of 30%, accounting for 80% of the city's total and one-fourth of the national total [3]. - Haidian District has the highest concentration of top AI talent and laboratory resources in China, supported by various government initiatives to foster AI development [3]. Group 2: Investment Timing and Strategy - Early investment in AI applications is deemed advantageous, with a focus on identifying when technologies will mature. The current period is likened to the mobile internet boom around 2011-2012 [4]. - Entrepreneurs are encouraged to act quickly once a direction is determined, as the market is rapidly evolving and the cost of market education is decreasing [5]. Group 3: Demand and Market Dynamics - Investors and entrepreneurs agree on the importance of distinguishing between genuine and artificial demand, advocating for solutions that enhance efficiency rather than creating unnecessary AI applications [7]. - The demand for AI applications is categorized into three types: cost reduction for businesses, new value experiences for individuals, and innovative human-computer interactions [8]. Group 4: Commercialization Challenges - There is a clear divide in opinions regarding whether to focus on B2B or B2C markets, with B2B models seen as more mature and having clearer customer payment logic [12]. - The challenges of monetizing C2C applications are highlighted, with a consensus that achieving product-market fit (PMF) is crucial for success [14]. Group 5: Globalization and Market Expansion - A notable trend is the early globalization of AI startups, with many companies choosing to target international markets from inception [16]. - Chinese companies are making significant strides in the global AI market, particularly in the field of embodied intelligence, with a focus on expanding overseas customer bases [18]. Group 6: Incubation Trends - Investment firms are increasingly engaging in incubation, with various models being adopted to support startups through funding and resources [20]. - The importance of exit strategies in the investment ecosystem is emphasized, with recommendations for entrepreneurs to align with industry funds for better resource access [21].

Summary of Key Points from the Conference Call Industry Overview - **Industry**: China Industrials - **Outlook**: The industrial sector in China is facing a challenging trajectory in 2H25, with persistent macro headwinds and a cautious outlook due to muted demand and external risks, particularly from US tariffs [10][11][24][25]. Core Insights 1. **Earnings Performance**: In 1H25, 39% of companies reported earnings beats, a notable increase from 20% in 2H24, indicating improved performance against lower expectations [1]. 2. **Manufacturing Activity**: The Manufacturing PMI fell below 50 during Apr-Aug 2025, reflecting weak domestic consumption and cooling export orders [11][12]. 3. **Corporate Profits**: Industrial profits declined by 1.7% year-on-year to RMB 4 trillion (approximately USD 559 billion) in 7M25, with a slight recovery noted in July due to government measures [14]. 4. **Capex Intentions**: There is a significant contraction in Japan's machine tool orders to China, indicating a risk-off sentiment among manufacturers [16][20]. 5. **Destocking Cycle**: The destocking phase is nearing an end, but restocking is not yet in sight, as businesses await improved demand and profit margins [21]. Investment Strategies Barbell Strategy - **High-Risk Basket**: Focus on sectors like AI infrastructure, factory automation, and humanoid robots. Key picks include: - **AI Infra**: Kingboard Laminates (KBL), Shengyi Technology (SYTECH), Han's CNC [26][27]. - **Factory Automation**: Wuxi Lead, UBTECH, Hengli Hydraulic [43][46]. - **Low-Risk Basket**: Emphasize infrastructure and export sectors, with a preference for: - **China Infrastructure**: CRRC, Lesso, China State Construction International (CSCI) [5][61]. - **Export**: Techtronic, Shenzhou, Stella, focusing on high dividend yields [5]. Key Company Insights 1. **Kingboard Laminates (KBL)**: Reported 1H25 earnings growth of 28% to HKD 933 million, with expectations of improved gross margins in 2H25 due to price increases [28][29]. 2. **Shengyi Technology (SYTECH)**: Anticipates a 10-15% increase in shipments of AI-related materials, with ongoing expansion plans [33][34]. 3. **Wuxi Lead**: Expected to benefit from an EV battery capex cycle turnaround, with new orders projected to exceed previous guidance [47][48]. 4. **UBTECH**: Revised delivery guidance for humanoid robots upwards, indicating strong demand in the auto and electronics sectors [52][53]. 5. **CRRC**: Upgraded to Buy due to strong earnings and increased high-speed rail tenders, with a target price raised to HKD 7.30 [62][64]. Additional Considerations - **Policy Response**: The effectiveness of government policies in stimulating demand remains uncertain, with a need for decisive action to restore private sector confidence [24]. - **Market Sentiment**: The overall sentiment in the industrial sector is cautious, with a preference for companies with strong balance sheets and exposure to structural growth themes [25]. This summary encapsulates the key points discussed in the conference call, highlighting the current state of the China industrial sector, investment strategies, and specific company insights.

Core Viewpoint - The demand for computing power infrastructure is expanding due to the explosion of AI model requirements and the intelligent transformation across various industries, creating multi-layered investment opportunities in cloud computing, AI+Data, AI Agents, and AI computing power [1] Company Performance - Oracle's first fiscal quarter results showed mixed performance with revenue of $14.93 billion (up 12% year-on-year, below the expected $15.03 billion) and adjusted earnings per share of $1.47 (slightly below the expected $1.48) [1] - Oracle's cloud business experienced strong growth, with total cloud revenue of $7.2 billion (up 28% in dollars, 27% at constant currency), including IaaS revenue of $3.3 billion (up 55%) and SaaS revenue of $3.8 billion (up 11%) [1] - The company expects its cloud infrastructure business to grow by 77% this fiscal year, reaching $18 billion, with projected revenues for the next four years significantly exceeding previous expectations [1] Remaining Performance Obligations - Oracle's remaining performance obligations reached $455 billion, a year-on-year increase of 359%, with a significant contract signed with OpenAI valued at approximately $30 billion [2] - The CEO indicated that new multi-billion dollar contracts are expected to be signed in the coming months, potentially pushing remaining performance obligations over $500 billion [2] Market Demand Trends - Coreweave is experiencing a surge in demand for computing power, with long-term contracts becoming the norm as clients' needs have expanded from thousands to millions of GPUs [3] - The company has integrated approximately 2.2 GW of capacity, with 900 MW expected to be operational by the end of the year, indicating a supply-demand imbalance in computing power infrastructure [3] Capital Expenditure Trends - Major global cloud service providers (CSPs) are increasing capital expenditures, with Microsoft, Google, Meta, and Amazon raising their spending forecasts for AI infrastructure and data center expansions [4] - Microsoft plans to increase its capital expenditure to over $30 billion in FY2026, while Google has raised its capex to $85 billion for 2025 [4] - In China, companies like Alibaba and Tencent are also ramping up investments in AI and cloud infrastructure, with Alibaba's CEO indicating that the next three years will see more investment than the past decade combined [5] Investment Targets - Suggested investment targets include companies in cloud computing such as Kingsoft Cloud, Alibaba, and Tencent [7] - For AI+Data, companies like StarRing Technology and DaMeng Data are highlighted [8] - In the AI Agent sector, companies like Dingjie Zhizhi and Vision Source are recommended [8] - For AI computing power, companies such as Cambricon and Inspur Information are noted as potential investment opportunities [9]

Group 1: Bending Spoons Acquisitions - Bending Spoons (BS) acquired Brightcove for $233 million in cash and subsequently privatized it [1] - BS has now acquired Vimeo for $1.38 billion in cash, indicating a shift towards acquiring larger, previously high-valued public companies [2][3] - The CEO of BS, Luca Ferrari, stated their intention to indefinitely own and operate acquired companies, with plans for significant investments in key markets and product areas post-acquisition [3] Group 2: Mercor's Growth in AI Recruitment - Mercor, an AI recruitment platform, achieved a valuation of $2 billion after a $100 million Series B funding round and has reached an annualized revenue of $450 million, reflecting a 4.5x growth in just six months [5] - Major clients of Mercor include Google, Amazon, Meta, Microsoft, OpenAI, and NVIDIA, positioning it as a strong competitor to Scale AI and Surge [5] - Mercor's CEO clarified that their annual recurring revenue (ARR) exceeds $450 million, although this figure is akin to gross merchandise volume (GMV) and does not account for downstream fees [6] Group 3: AI Industry Trends - The AI industry is experiencing explosive growth, as evidenced by the rapid revenue increases of companies like Mercor and others in the AI coding sector [7] - There is ongoing debate within the industry regarding revenue calculation methods, particularly concerning the costs associated with large models [7] - Another AI infrastructure product has seen a 40-fold increase in monthly revenue, indicating a strong demand for AI applications and services [8]