2025年算力调度平台行业 优化计算资源，支撑AI应用 ◼ 研究背景 随着人工智能技术的迅猛发展，全球 范围内对算力的需求呈现指数级增长， 因此需要算力调度来实现跨地域、跨 平台的算力资源整合与优化调度。 ◼ 研究目标 概览标签：算力、算力调度 Computing Power Scheduling Platform Industry 計算力スケジューリングプラットフォーム業界 报告提供的任何内容（包括但不限于数据、文本、图表、图像等）均系头豹研究院独有的高度机密性文件（在报告中另 行标明出处者除外）。 ，任何人不得以任何方式擅自复制、再造、传播、出版、引用、改 编、汇编本报告内容，若有违反上述约定的行为发生，头豹研究院保留采取法律措施、追究相关人员责任的权利。头豹 研究院开展的所有商业活动均使用"头豹研究院"或"头豹"的商号、商标，头豹研究院无任何前述名称之外的其他分支机构 ，也未授权或聘用其他任何第三方代表头豹研究院开展商业活动。 头豹研究院 1 01 异构算力调度面临的挑战： ◆ 异构算力调度面临多重核心挑战：资源异 构性与软件环境碎片化显著增加调度复杂 性；跨架构任务迁移成本高导致效率低下； 缺乏统一调 ...

周红波介绍了南京在未来网络领域取得的突破性进展，他表示南京将坚持走求实扎实的创新路子，与各 方一道，面向前沿、打造科研攻关总平台，突出融合、建好未来产业先导区，聚焦应用、构建开放共享 大网络，共建生态、拓展全链赋能新空间，加快建设全球未来网络创新枢纽。 大会围绕算网融合、智算网络、网络大模型、确定性网络等热点话题，举办主论坛、18场分论坛，以及 未来网络领域创新成果发布、未来网络科技创新大赛、创新成果展示等活动。国家相关部门和省、市有 关方面负责同志，未来网络领域专家、行业企业代表，高校院所和部分其他省市代表参加。 本报讯(记者张宣蔡姝雯杨易臻)8月20日，第九届未来网络发展大会在南京开幕。本届大会以"网络全球 决胜未来"为主题，现场发布未来网络领域最新重大科技成果，包括全域一体化数算网调度平台、行业 首个面向广域网的网络大模型与多智能体系统、基带电路AI自动设计工具链等。国家数据局党组成 员、副局长夏冰，省委常委、南京市委书记周红波，副省长赵岩出席开幕式。 夏冰表示，设施互联互通、"网—算—数"协同，离不开先进网络的支撑。希望未来网络持续锻造高可靠 优势，着力构建定制化能力，瞄准产业应用场景，充分释放融合应 ...

Investment Rating - The industry investment rating is "Outperform the Market," indicating that the industry index is expected to perform better than the benchmark index over the next 6-12 months [11]. Core Insights - The report highlights that the three major telecom operators in China—China Mobile, China Telecom, and China Unicom—are transitioning from traditional communication providers to "AI infrastructure providers, industry enablers, and security guardians," significantly contributing to the digital transformation of the economy and society [1][3]. - The integration of AI into various sectors is seen as a key driver of the new technological revolution and industrial transformation, with telecom operators playing a crucial role in this evolution [1]. Summary by Sections Investment Recommendations - The report suggests that telecom operators will significantly benefit from the performance growth and value reassessment brought by intelligent computing services under the AI wave. It recommends focusing on the three major operators: China Mobile, China Telecom, and China Unicom, as well as their partnerships with computing service providers such as Runze Technology, Pingzhi Information, Zhongbei Communication, ZTE, Fenghuo Communication, and Unisplendour [3]. Industry Developments - The report emphasizes the practical AI capabilities demonstrated by the telecom operators at WAIC 2025, showcasing their role as a "digital infrastructure national team" and their technological leadership in building a smart service ecosystem that covers both industry and people's livelihoods [3]. - The collaboration between computing power and network innovation is highlighted as a core logic for the operators' computing infrastructure, leading to significant breakthroughs in computing resource utilization and efficiency [3]. Infrastructure Investment - In 2025, the three major operators are expected to increase their investments in computing infrastructure, with China Mobile planning to invest 37.3 billion RMB in this area. Both China Telecom and China Unicom are projected to see over 20% year-on-year growth in their computing investments [3]. - The report notes that the domestic server procurement by China Unicom has exceeded 90% for domestic servers, while China Mobile has the largest single liquid-cooled intelligent computing center among global operators, with over 85% of its computing power being domestically sourced [3].

Core Insights - The report highlights the ongoing development and commercialization of humanoid robots, supported by industrial funds for innovative breakthroughs, with related stocks including Zhongdali De (002896) and Industrial Fulian (601138) [8] - The implementation of stablecoin regulations in Hong Kong is accelerating the global layout of digital financial infrastructure, with related stocks such as Sifang Jingchuang (300468) and Hengsheng Electronics (600570) [8] - The first intelligent storage scheduling platform in the country has been launched to promote the integration of computing and networking, with related stocks including Unisplendour Corporation (000938) and Zhongke Shuguang (603019) [8] Company Focus - Redick (300652) has completed the development of planetary roller screw products, covering all types and achieving mass production [10] - Jinyi Culture (002721) has collaborated with several banks on digital RMB products through its subsidiary Kaike Weishi, and is also involved in AI large model-related projects [10] - Wantong Intelligent Control (300643) has signed an exclusive sales and cooperation contract with Shanghai Shenming Aosi Semiconductor Technology Co., Ltd. for the production and sale of boards based on the Fellow 1 model chip [10] Market Review and Outlook - On July 30, the total trading volume in the two markets was 1,844.3 billion yuan, with 1,632 stocks rising and 3,376 stocks falling, indicating an increase of 41.1 billion yuan compared to the previous day [12] - The major indices opened flat and experienced wide fluctuations, with the exception of the Shanghai Composite Index, which closed slightly higher, while other important indices closed with small losses [12] - Market hotspots included significant gains in sectors such as film, oil and gas, and dairy, while sectors like digital currency and diversified finance saw declines [12][13]

Core Insights - The integration of AI technology is becoming a core driver for industrial transformation, with significant opportunities estimated in the trillions [1][2] - The application of 5G technology is accelerating in various sectors, contributing to the digital transformation of industries [3] Group 1: AI and Industrial Transformation - A large-scale industrial model has produced 25,000 industrial models and empowered over 75 million 5G factories, showcasing the extensive application of AI [1] - AI-native applications have surpassed 270 million monthly active users, and devices with AI capabilities have exceeded 1.4 billion [1][2] - The "AI Full Integration" business incubation center was launched, indicating a strategic focus on AI as a foundational element in IT infrastructure [1][2] Group 2: 5G Technology and Manufacturing - 5G technology has been integrated into 86 out of 97 categories of the national economy, facilitating digital transformation across various industries [3] - The industrial internet sector has reached a scale of 1.5 trillion yuan, with over 18,500 industrial 5G projects established nationwide [3] - In the aircraft manufacturing sector, 5G has improved equipment utilization rates to 90% and reduced inspection times by 80% [2] Group 3: Collaborative Innovation and Research - A joint innovation platform was established to focus on areas such as network security, AI, and quantum technology, promoting collaborative research and application [3] - The Fudan University team has developed a new tactile sensor with a high density of sensing units, enhancing robotic capabilities [4]

Core Viewpoint - China Unicom Cloud aims to establish itself as the "first brand of computing power" in the context of national strategy, emphasizing the importance of self-controlled computing power for national strength [1][6]. Group 1: Strategic Positioning - The strategic positioning of China Unicom Cloud is driven by the national consideration of "computing power equals national power," highlighting its responsibility to meet the country's demand for self-controlled computing capabilities [1]. - The company has made significant investments in core underlying technologies, such as the domestically developed container engine CSK Turbo, which addresses the issue of dependency on foreign open-source technologies [1][2]. Group 2: Technical Innovations - China Unicom Cloud has achieved a scale of 400,000 real-time online containers internally, with a single server supporting over 1,000 Pods and a 70% reduction in memory usage [2]. - The company has developed an advanced computing power scheduling platform called "Unicom Star Network," which enables unified orchestration and hierarchical scheduling of computing resources across the network [3]. Group 3: Market Solutions - To address the supply-demand imbalance in the current agent market, China Unicom Cloud has launched a "Digital Product Supermarket," aiming to create a more open and transparent market for intelligent agents [3][4]. - The "Digital Product Supermarket" offers over a hundred selected intelligent agents categorized by different types, catering to the diverse needs of enterprise users while ensuring security [4][5]. Group 4: Competitive Advantages - China Unicom Cloud differentiates itself through a strong security philosophy, developing eight internal security products that provide proactive security measures [6]. - The company boasts unparalleled localized service capabilities, with marketing, delivery, and support reaching down to the county level, which is a significant advantage over other cloud service providers [6].

Core Viewpoint - The commercialization of 5G technology is accelerating, with 5G-A evolving to provide faster speeds, lower latency, and broader connectivity, significantly transforming lifestyles and industries, powered by the rapid development of AI technology [1][11]. Group 1: 5G-A Deployment and Investment - China's three major telecom operators have completed 5G-A coverage in 31 provinces and over 300 cities, focusing on key areas such as business districts and transportation hubs [1]. - China Mobile plans to invest 9.8 billion yuan to open over 100,000 tri-band aggregation base stations within the year [1]. - China Unicom has launched 5G-A services in 39 major urban areas and is conducting low-altitude communication demonstrations across 31 provinces [1]. - China Telecom has deployed approximately 75,000 5G-A base stations, covering 121 cities [1]. Group 2: Smart Home Innovations - Shanghai Unicom has introduced "Smart Home Assistant," an AI robot that integrates multiple functions such as video calls, digital archives, and home monitoring, enhancing convenience through voice interaction [3][7]. - The AI camera for infant monitoring features advanced capabilities like face covering detection and real-time alerts to parents, addressing traditional security blind spots [7]. - The "Smart Home Assistant" also supports educational content for children and fitness programs for the elderly, effectively acting as a personal trainer [7]. Group 3: Network Performance and User Experience - In a performance test, 5G-A mobile download speeds reached 9389.45 Mbps, while fixed network speeds were recorded at 1191.08 Mbps with a ping of only 2 milliseconds, allowing for rapid downloads of large files [9]. - The low latency of the network supports seamless remote collaboration and smart device control, ensuring stable connections for multiple devices [9]. Group 4: Intelligent Transportation Systems - The "Double 20" vehicle networking demonstration zone showcases a significant technological breakthrough with a 20 ms end-to-end latency and 20 Mbps upload speed, essential for future autonomous driving applications [10]. - The system can provide real-time traffic data to optimize public transport schedules and improve road efficiency by over 30% [10]. - Collaborations between Shanghai Telecom and public transport groups have resulted in increased public transport usage and reduced congestion [10]. Group 5: Future Outlook - The integration of 5G-A and AI is transforming communication technology from mere connectivity to a core driver of smart living and industrial change, shaping a digital future for China [11]. - The convergence of computing and networking technologies is making advanced technology more accessible and impactful in everyday life [11].

Core Insights - Nvidia reported record revenue of 441 billion yuan for Q1 of fiscal year 2026, exceeding expectations in a competitive global AI landscape [1] Financial Performance - Data center revenue reached 39.1 billion USD in Q1, marking a 10% increase from the previous quarter [2] - Nearly 70% of data center revenue was contributed by the new Blackwell architecture GPUs, as the transition from Hopper to Blackwell nears completion [2] Market Trends - AI is becoming a new global infrastructure, with a focus on sovereign AI and enterprise-level AI agents [2] - The reshaping of the global landscape is driving countries to relocate manufacturing, creating massive data and computational needs [2] Technological Developments - The integration of computing power networks is expected to foster innovative applications in various sectors, including smart mining and smart cities [3] - The demand for computing power is critical for the implementation of AI across industries, with a focus on ease of use, flexibility, and scalability [3] Company Initiatives - WIMI is advancing its AI computing strategy by building a high-end computing base and collaborating with industry partners to enhance factory efficiency [4] - WIMI's holographic cloud platform supports large model training and inference, aiming to reduce latency and improve energy efficiency [4] Industry Outlook - Nvidia is positioned as the largest chip company globally by revenue and market capitalization, leading the AI computing infrastructure sector [5] - The diverse range of hardware and software products from Nvidia is enabling larger-scale sales to customers, indicating a significant shift in the global tech landscape [5]

Core Viewpoint - The report emphasizes that computing networks, as a new type of infrastructure combining "computing + connectivity," are crucial for the digital transformation of the industrial internet, aiming to achieve ubiquitous computing, symbiotic networks, intelligent orchestration, and integrated services [1][19]. Group 1: Development Background and Trends - The digital economy and industrial digital transformation require superior network connectivity and robust computing power to meet demands for ultra-high bandwidth, ultra-low latency, and high security [20]. - Computing networks represent a fusion of new infrastructure demands, integrating 5G, industrial internet, AI, cloud/edge computing, and blockchain technologies [21]. - The development of computing networks is driven by the convergence of 5G and edge computing technologies, enhancing service reliability and reducing latency [21][22]. Group 2: Current Status of Computing Networks - Computing networks have become a hot topic in the industry, with operators, equipment manufacturers, and research institutions actively participating in research and establishing a development roadmap and architecture [22][23]. - Initial achievements have been made in core technology areas such as computing modeling, perception, routing, and trading, with various standards being developed [23][24]. Group 3: Application Scenarios of Computing Networks - Computing networks will transform the supply, application, and service methods of computing power, enhancing service flexibility and efficiency across various industries [27]. - Specific applications include: - **Cloud VR Video Services**: Utilizing multi-layered computing power to meet the increasing demand for immersive experiences [28][29]. - **V2X Vehicle Networking**: Integrating AI and IoT to enhance data processing and transmission capabilities for smart transportation [30][34]. - **Green and Low-Carbon Initiatives**: Addressing energy consumption issues in data centers by optimizing resource utilization and supporting carbon neutrality goals [33][36]. Group 4: Key Technologies in Computing Networks - Key technologies include unified resource description frameworks, IT/OT data boundary integration, multi-factor routing based on SRv6/IPv6, and digital twin networks for optimizing industrial operations [2][19]. - The focus is on overcoming technical bottlenecks related to heterogeneous hardware abstraction and generalization of computing primitives to support deep integration in industrial scenarios [2]. Group 5: Vision and Future Outlook - The vision for computing networks is to achieve a state where computing power is as accessible as utilities like water and electricity, enabling seamless integration of services across various sectors [21][22]. - Future developments will focus on enhancing industry collaboration and deepening the integration of computing networks in industrial applications, positioning them as a core engine for digital economic growth [2].

Core Viewpoint - The report emphasizes the integration of computing networks and industrial internet, highlighting that computing networks serve as a new type of information infrastructure centered on computing and rooted in networking, facilitating the digital transformation of industries [1][21]. Group 1: Development Background and Trends - The digital economy and industrial digital transformation require superior network connectivity and robust computing power to meet demands for ultra-high bandwidth, ultra-low latency, and high security [22]. - Computing networks are emerging as a hot topic in the industry, with operators, equipment manufacturers, and research institutions actively engaging in research and establishing a development roadmap and architecture system [24][25]. - International organizations like ITU and IETF are working on establishing standards for computing networks, indicating a global push towards computing and network convergence [26][27]. Group 2: Key Technologies - Key technologies for computing networks include unified metrics for computing and network, real-time user demand and state perception, multi-factor joint scheduling algorithms based on IPv6/SRv6, and ensuring end-to-end service reliability [2]. - The report identifies the need to overcome technical bottlenecks such as heterogeneous hardware abstraction and edge computing collaboration to optimize industrial production processes [2]. Group 3: Application Scenarios - Computing networks will transform the supply, application, and service methods of computing power, enhancing service flexibility and efficiency across various industries [29]. - Specific application scenarios include Cloud VR services, V2X vehicle networking, and green low-carbon initiatives, all benefiting from the integration of computing and networking capabilities [30][32][37]. Group 4: Collaborative Development - The industrial internet network is crucial for connecting people, machines, and objects in industrial environments, with internal and external networks serving different operational needs [39].