Core Insights - The 2025 China International Information and Communication Exhibition highlighted the critical phase of deep integration between all-optical capacity and computing networks, emphasizing "low latency, high reliability, and wide adaptability" as industry consensus [1][2] - The demand for computing power is surging due to various digital applications, making all-optical capacity a fundamental support for the development of the digital economy [1] - The goal of current all-optical capacity development is to promote continuous innovation and facilitate the deep integration of computing power across industries [1] Industry Developments - The three major telecom operators in China are transitioning from "network deployment" to "network optimization," with China Unicom developing "Computing Power Intelligent Network AINet," China Mobile upgrading its "Jiuzhou" computing optical network, and China Telecom advancing its All-Optical Network 3.0 [2] - China Telecom's All-Optical Network 3.0 strategy aims for deep integration of optical networks and cloud resources, creating an efficient and intensive all-optical connection base for computing, storage, and transport resources [2] Technological Innovations - "Optical Intelligence Dual Empowerment" is identified as a key direction for overcoming the bottlenecks in the integration of computing and networking [2] - Huawei's AI-OTN focuses on "Optical Intelligence Integration" to establish a millisecond-level capacity base, enabling on-demand access to computing power and integrating intelligent technologies into optical networks for advanced self-intelligence evolution [2] Reports and Standards - The China Academy of Information and Communications Technology released several reports, including the "Computing Network Carrying Capacity Index Assessment Report (2025)" and the "High-Quality All-Optical Computing Power Dedicated Line Research Report (2025)" [3] - The reports emphasize actions to promote millisecond interconnection of computing centers, millisecond access to computing resources, and millisecond availability of computing applications, enhancing the end-to-end connectivity capabilities of networks [3]

本报记者 梁傲男 近日，在新一代计算标准工作委员会主办的"新一代计算产业大会"上，中科驭数（北京）科技有限公司（以下简称"中科 驭数"）CEO鄢贵海带来题为《数据处理器（DPU）构建新一代AI网络的优势与关键技术》的主题报告。 "之前我们做过一个比喻：如果把CPU比做大脑、那么GPU就好比是肌肉，DPU是神经中枢。CPU承载了应用生态，决定 了计算系统是否可以通用；GPU提供了大量不同精度要求的算力，决定了系统是否有足够的'力量'；DPU负责数据在各种CPU 和GPU之间高效流通，决定了系统是否能协同工作。"鄢贵海表示。 他认为，随着人工智能的快速发展，传统计算架构已经无法满足海量数据的处理需求，基于DPU高速网络的新一代计算架 构成为提升效能的关键。DPU通过卸载GPU的计算负担、优化数据预处理、加速通信和存储任务，正成为大模型训练和推理的 重要助力。 四是面向AI集群的后端网络网卡，特点是支持Scale-Out的高性能RDMA网卡和支持Scale-Up的高性能网卡，前者兼容RoCE 协议，后者将支持基于各种GPU构造算力的超节点。当前主力产品是FlexFlow 2200R。 据记者了解，以上产品传输数据带 ...

Core Viewpoint - The announcement by Huigreen Ecological Technology Group regarding the lease agreement for a photonic industry park reflects a broader trend of capacity expansion in the optical module industry, driven by surging AI computing demands [1] Industry Expansion - The optical module industry is experiencing a wave of capacity expansion, with major companies focusing on technological upgrades and production capacity to seize market opportunities [2][3] - Huigreen Ecological has signed contracts to build production bases for 450 million optical modules with a total investment of 700 million yuan, indicating significant growth in production scale and technology [2] - Other companies like Shanghai Cambridge Technology and Zhongji Xuchuang are also ramping up production of high-end products, including 1.6T and 800G optical modules, to meet rising demand [2] Market Demand and Growth - The demand for 400G and 800G optical modules remains strong, with the 1.6T optical module expected to become a key growth driver by 2025 [2] - The global optical module market is projected to grow from $11.2 billion in 2020 to $17.8 billion in 2024, with a compound annual growth rate of 12.2% [3] - In China, the optical module market is expected to reach approximately 70 billion yuan by 2025, reflecting a growth of 12.22% from the previous year [3] Performance Metrics - Leading domestic optical module companies have reported significant sales growth, with Xinyi Sheng's sales volume reaching 6.95 million units in the first half of 2025, a year-on-year increase of 112.5% [4] - Huagong Technology reported a revenue of 3.744 billion yuan from its connectivity business in the first half of 2025, a 124% increase year-on-year, with expectations of substantial growth in AI-related optical module shipments [4] Supporting Factors - The growth in the optical module industry is supported by a surge in AI computing infrastructure both domestically and internationally, with major tech companies increasing their procurement of high-speed optical modules [5] - Policy support from the Chinese government, including guidelines for enhancing data infrastructure, is expected to further stimulate the optical module market [5] - The long-term opportunities in the optical module industry are driven by the continuous iteration of AI technology, expanding demand, and supportive policies [5]

Group 1 - The Jiangsu Industry-University-Research Cooperation Conference aims to promote deep integration of technological and industrial innovation, serving as a platform for high-end innovation resource aggregation and technology achievement transformation [1][6] - The conference featured the "Kirin Team" from the Nanjing Kirin Science and Technology Innovation Park, showcasing multiple latest achievements from five research institutes affiliated with the Chinese Academy of Sciences, indicating the growing influence of national strategic scientific forces [1][2] - The theme of this year's conference is "Technology Empowering New Quality Productivity, Innovation and Integration Development Leading the Way," with various activities organized to facilitate deep connections among the technology, industry, and finance sectors [1][6] Group 2 - The Nanjing Mobile Communication and Computing Innovation Research Institute presented a comprehensive system equipment platform for integrated information networks, achieving breakthroughs in low-orbit satellite communication technology [2][3] - The research institute has successfully incubated nearly 30 satellite communication-related enterprises, including 10 high-tech companies, focusing on satellite communication and low-altitude self-organizing networks [2][3] - The "Information High-speed Rail Comprehensive Experimental Infrastructure" was highlighted as a significant research facility, aimed at providing a comprehensive experimental environment for the integration of computing and networking [2][3] Group 3 - The Nanjing Information High-speed Rail Research Institute showcased two key products: an AI development platform that supports over 50 major scientific research models and a model factory service platform for AI integration [3][4] - The Future Energy Systems Research Institute presented advanced energy storage technologies, including a 100MW advanced compressed air energy storage system, which is the first of its kind with complete independent intellectual property rights [3][4] Group 4 - The Nanjing Software Technology Research Institute introduced a major open-source software infrastructure, which has become the largest open-source software governance platform in China, supporting over 80 leading enterprises [4][5] - The Nanjing Life Health Research Institute showcased advanced products in the fields of innovative vaccines and high-end medical devices, including a broad-spectrum influenza vaccine with a unique design [5][6] Group 5 - The Nanjing Kirin Science and Technology Innovation Park has become a significant hub for innovation resources outside Beijing, with multiple major scientific and technological infrastructure projects focusing on critical core technologies [5][6] - The park is actively implementing actions to build an industrial technology innovation highland, aiming to convert high-end innovation resources into high-quality development momentum [6]

Group 1 - The white paper published by China Unicom Research Institute in August 2025 focuses on the integration of computing and networking (算网融合) CDN technology, analyzing the research background, development status of Unicom's CDN, key technologies, and future directions to provide industry references [1][5][8] - The Internet 2030 initiative promotes the construction of new digital infrastructure, with the integration of computing and networking as a key feature that empowers CDN development, especially with the increasing demand for differentiated services in emerging video applications [1][18] - As of May 2024, the proportion of active IPv6 users in China reached 72.70%, with significant improvements in the deployment of IPv6 nodes by CDN companies, laying a foundation for network development through IPv6+ technology [1][31][32] Group 2 - China Unicom's CDN development includes the launch of a new self-developed CDN product in July 2024, with nodes deployed in core data centers, covering all provinces and serving top enterprises [1][39] - The intelligent CDN development is planned in three stages: 1.0 focuses on building data center CDN; 2.0 utilizes MEC and BNC to develop edge CDN; 3.0 aims to integrate AI edge inference with CDN [1][40][41] - The edge CDN is categorized into three types: fixed network edge CDN, mobile network edge CDN, and integrated edge CDN, enhancing user experience through node sinking, intelligent scheduling, and tiered guarantees [1][45][48] Group 3 - Key technologies for the integration of computing and networking CDN include SRv6 routing optimization, application-aware CDN, and unified transport protocols, which enhance efficiency and reduce latency [2][29] - The SRv6 technology can reduce routing adjustment periods from three minutes to one second, achieving a latency reduction of 20% to 30% [2][29] - The application-aware CDN, based on APN technology, breaks the traditional separation between CDN and network, providing acceleration services at three "Mile" points to address operational pain points for telecom networks [2][29] Group 4 - The demand for bandwidth is expected to grow exponentially due to the rise of ultra-high-definition video applications, with 4K video requiring bandwidth between 22.5M to 75M and 8K video requiring 90M to 300M [1][43][44] - The rapid growth of internet data traffic, particularly driven by video services, necessitates the development of edge CDN solutions to alleviate pressure on backbone networks [1][44] - China Unicom aims to leverage its MEC operations to develop intelligent CDN solutions that support ultra-high-definition and low-latency content distribution [1][44][52]

Core Insights - The report titled "China Unicom's Computing and Network Integration CDN Technology Research White Paper (2025)" focuses on the development and application of CDN technology under the trend of computing and network integration, addressing the bandwidth and low-latency demands of emerging video applications such as ultra-high-definition 3D live streaming, XR, and cloud gaming [1][12]. Group 1: Research Background - "Internet 2030" outlines six key features, including full interconnectivity and computing-network integration, emphasizing the need to break down barriers between computing, data, and networks [1][17]. - CDN has evolved through three stages: traditional transmission, content distribution, and integration with network operations, necessitating collaboration with underlying networks for effective computing-network integration [1][22]. - As of May 2024, there are 794 million active IPv6 users in China, with 96.14% of typical CDN companies having completed IPv6 node upgrades, laying a foundation for CDN advancements [1][26][28]. Group 2: China Unicom CDN - Cloud Universe Technology Co., Ltd., a joint venture between China Unicom and Tencent, focuses on CDN and edge computing, supporting the construction and development of China Unicom's CDN network [1][32]. - China Unicom's CDN network has been established nationwide, serving top domestic internet companies and mobile manufacturers, with plans to expand to cover all prefecture-level cities within three years [1][35]. - The development of China Unicom's intelligent CDN is planned in three phases: 1.0 focuses on traditional content distribution, 2.0 leverages MEC and BNC for low-latency video services, and 3.0 aims to integrate AI edge inference with CDN and MEC for real-time content distribution [1][38]. Group 3: Key Technologies in Computing-Network Integration CDN - SRv6 routing optimization reduces dynamic acceleration latency by 20%-30% by slicing forwarding paths and constructing low-latency acceleration channels [1][25]. - Application-aware CDN utilizes APN technology to deeply integrate business distribution and network scheduling, addressing the traditional disconnect between CDN and networks [1][24]. - Unified transport protocols focus on QUIC and MoQ, providing standards for AIGC content distribution, supporting low latency and publish/subscribe models [1][29].

Core Viewpoint - The Ministry of Industry and Information Technology and the National Forestry and Grassland Administration have jointly issued a notice to promote the construction of "Broadband Forestry and Grassland," aiming for significant improvements in network coverage in forest and grassland areas by the end of 2027 [1][2]. Group 1: Key Objectives and Tasks - By the end of 2027, 90% of forest station areas are expected to have 4G/5G network coverage, with significant enhancements in key locations such as populated areas and fire lookout towers [1]. - The notice outlines nine tasks focusing on enhancing network coverage in important locations, strengthening network construction in key areas, optimizing the construction environment, and improving network maintenance and usage levels [2]. Group 2: Communication Infrastructure Development - The notice emphasizes the need for various communication transmission methods, including fiber optics, microwave, and satellite, to optimize network coverage in natural protected areas [3]. - It highlights the importance of collaboration between telecommunications companies and local governments to ensure funding and support for the "Broadband Forestry and Grassland" projects [4]. Group 3: Current Achievements and Future Plans - As of July 2023, China has established 4.598 million 5G base stations and over 30.53 million gigabit network ports, achieving "gigabit access in every county and 5G access in every township" [4]. - The Ministry of Industry and Information Technology plans to continue enhancing network construction and application, focusing on 5G and gigabit optical networks, and advancing research and development in 6G technology [5].

Core Viewpoint - The article highlights the advancements in China's new information infrastructure, particularly in 5G and computing power, showcasing significant upgrades in coverage, efficiency, resource sharing, and connectivity [1][2][3]. Group 1: Infrastructure Development - As of July 2023, China has built 4.598 million 5G base stations and over 30.53 million gigabit network capability ports, achieving "gigabit access in every county and 5G access in every township" [1]. - A special action for "signal enhancement" is being implemented to deepen coverage of 5G and gigabit optical networks in key areas such as cultural tourism and transportation hubs, improving network quality and user experience [1]. Group 2: Computing Power Integration - The construction of computing power infrastructure is progressing steadily, with 10.85 million standard racks in use and an intelligent computing power scale reaching 788 EFLOPS (FP16 half precision) as of June 2025 [2]. - The national integrated computing power network system is being accelerated, with transmission delays between national hubs not exceeding 20 ms and between regional computing nodes not exceeding 1 ms [2]. Group 3: Resource Sharing and Sustainability - The sharing rate of newly built towers has reached 85%, with over 95% of new 5G base stations utilizing existing site resources, equivalent to saving 1.2 million communication towers and 68,000 acres of land [2]. - The promotion of 5G inter-network roaming has effectively reduced redundant investments and enhanced the overall efficiency of the 5G network [2]. Group 4: Rural and International Connectivity - Thanks to ten years of continuous construction of universal telecom services, rural broadband network development has accelerated, achieving "broadband access in every administrative village" and over 90% 5G coverage [3]. - As of now, there are 15 international gateway exchanges, and telecom companies have invested in 17 international and Hong Kong-Macau-Taiwan submarine cable systems, significantly enhancing international communication capabilities [3].

Core Insights - Shenzhen Airport's T2 terminal and related projects have officially commenced construction, aiming to enhance the airport's capacity and efficiency, supporting its development as a composite international aviation hub [1][2] - The design capacity for the new projects includes an annual passenger throughput of 31 million, with significant infrastructure developments such as a 400,000 square meter terminal and various transport centers [1] - By 2024, Shenzhen Airport is projected to exceed 60 million passengers, ranking fourth nationally, with a cargo throughput of 188,100 tons, maintaining the top position for domestic cargo [1][3] Infrastructure Development - The new T2 terminal and associated projects will significantly increase Shenzhen Airport's operational capacity, facilitating better integration with existing facilities [1][2] - The airport's expansion plans include three runways and multiple terminals, aiming to create a comprehensive transportation hub that integrates air, land, and rail [2] Future Projections - By 2035, Shenzhen Airport aims to achieve an annual passenger throughput of 100 million and a cargo throughput of 350,000 tons, with a target of over 120 international routes [3] Cargo Performance - Shenzhen Airport ranked 19th globally in cargo volume, handling 1,881,468 tons, reflecting a year-on-year growth of 17.6% [4][5] - The Greater Bay Area's airports, including Shenzhen, Hong Kong, and Guangzhou, are recognized among the top 20 global cargo airports, highlighting the region's strategic importance in global supply chains [4][7]

Investment Rating - The report does not explicitly provide an investment rating for the computing power scheduling platform industry. Core Insights - The rapid development of artificial intelligence technology has led to an exponential increase in global demand for computing power, necessitating computing power scheduling for resource integration and optimization across regions and platforms [2]. Summary by Sections Overview of the Computing Power Scheduling Industry - Computing power is defined as the ability of computer devices or data centers to process information, categorized into general computing power, intelligent computing power, and supercomputing power [15][18]. - China's computing power scale has grown rapidly, reaching 280 EFLOPS by 2024, with intelligent computing power accounting for 32% [20][23]. Challenges in Heterogeneous Computing Power Scheduling - Heterogeneous computing power scheduling faces multiple core challenges, including increased scheduling complexity due to resource heterogeneity and software environment fragmentation, high migration costs for cross-architecture tasks, and a lack of unified scheduling standards leading to resource mismatch and low utilization [4][43]. Major Domestic Computing Power Scheduling Platforms - National-level computing power scheduling platforms are primarily government-led or constructed by major operators, emphasizing cross-regional collaboration and market-oriented transactions [5][48]. - Provincial platforms cover key regions like the Yangtze River Delta and Chengdu-Chongqing, while municipal platforms focus on local AI and smart manufacturing scenarios [48]. Mainstream Open Source Computing Power Scheduling Technologies - Domestic computing power scheduling platforms are often built on open-source technologies, with openFuyao emerging as a versatile scheduling platform with advantages in domestic adaptation, while Kubernetes and Slurm have strong foundations in cloud-native and HPC fields [6][51].