前言：通信行业智能体的发展态势 通信网络目前正经历从 4G/5G/5G-A 向 6G 的 代际演进，万物互联的崭新时代已然开启。在赋能工 业互联网、智能电网、智能交通等垂直行业，提供强 大的超高速、可靠、实时通信保障，帮助众多企业提 升生产效率与产品质量的同时，通信网络自身运营和 运维，也面临运维效率、用户体验等关键挑战，通过 AI 和智能体技术注入，实现网络自身的智能化、自动 化技术转型，走向网络高阶自治，已成为产业共识。 业务和架构层面，2025 年 6 月，TM Forum 联 合 75 家全球领先的产业伙伴，共同发布了具有里程 碑意义的自智网络白皮书 7.0，定义了以自智网络智 能体（AN Agent）为核心的自智网络 Agentic 功能 架构和智能体架构，推动智能体、数字孪生技术与通 信网络深度融合。TM Forum 同时围绕自智网络 L4 等级发布了 20 个高价值场景的目标态与智能体能力， 以 KBI/KEI 指标量化商业价值，并借助 IG150x 系列 解决方案包，为行业提供部署指南，助力运营商量化 成效、提升部署效率。3GPP、ETSI、IETF、CCSA 等专业领域标准组织则聚焦落地使能 ...

根据公告，本次发行对象不超过35名合格投资者，发行股票数量不超过3120万股，占发行前公司总股本的30%，发行价格通过询 价方式确定，不低于定价基准日前20个交易日股票交易均价的80%。发行完成后，所认购股份限售期为6个月，公司控股股东王 飞雪、金建林、袁隽合计持股比例仍将保持绝对优势，控制权不会发生变化。 募集资金具体用途方面，2.73亿元将投入OSS产品研发平台及自智网络关键应用项目，通过构建统一研发基座升级15款现有OSS 产品，提升自智网络运维自动化与智能化水平；1.6亿元用于算力网络智能调度管理及运营系统项目，开发算力控制器、算网大脑 等9大产品模块，支撑"东数西算"战略实施；7400万元投向新一代数字家庭软件定义终端和管理平台项目，引入SDN（软件定义 网络）技术实现家庭网关转发与控制分离，助力电信运营商数字家庭业务升级；1.21亿元用于上海研发办公中心建设，改善研发 办公环境并提升人才吸引力；剩余3400万元补充流动资金，优化资本结构并支撑业务扩张。 根据募集说明书（修订稿），公司拟通过定向发行股票募集资金不超过6.62亿元，聚焦OSS产品研发平台及自智网络关键应用、 算力网络智能调度管理及运营 ...

Investment Rating - The report does not explicitly state an investment rating for the industry Core Insights - The rapid development of artificial intelligence is profoundly changing production and lifestyle, driving the upgrade of 5G core networks to AI core networks. Leading operators are actively practicing the commercial monetization of 5G-A intelligent core networks and considering the evolution towards an intelligent native Agentic Core [7][21] - The report identifies six major trends in the evolution of core networks, which have generated significant impact in the industry [7] Summary by Sections Trend One: Experience Management - The exploration of experience management for commercial monetization is beginning to show results, with operators enhancing operational capabilities through intelligence. Over 60% of users are willing to pay for better experiences, making experience rights a core competitive trend for operators [9][30] - Leading operators are exploring how to convert network experience into profitable rights products, combining high-end packages with scenario-based acceleration services to enhance user loyalty and open new revenue channels [30][31] Trend Two: AI Calling - AI Calling helps operators build an exceptional basic calling experience, evolving from voice assistants to multi-modal general assistants, thus seizing the AI business entry point. Operators are leveraging their unique advantages, such as a large user base and trusted numbering systems, to create differentiated competitive barriers [10][60] - The integration of AI capabilities into the entire calling process is key to enhancing user stickiness and stabilizing average call duration [57][61] Trend Three: Autonomous Network Level 4 (AN L4) - AN L4 has become an industry consensus, with network stability and operational efficiency being core strategic goals for operators. The TM Forum has designated 2025 as the commercial year for AN L4, with leading operators beginning to formulate their strategies [11][69] - The introduction of "digital employees" based on intelligent agent technology is playing a crucial role in complaint handling and alarm processing workflows [11] Trend Four: Intelligent Private Networks - The evolution of 5G-A connectivity capabilities is accelerating the development of intelligent private networks, which serve as accelerators for digital transformation across various industries. The integration of AI capabilities into connectivity is essential for building a robust foundation for enterprise AI applications [12] Trend Five: Collaborative Network Operations - The integration of communication, sensing, and computing capabilities is expected to release suppressed traffic and activate intelligent computing services. The report highlights the growing demand for high-definition video and the need for operators to enhance their network capabilities to meet this demand [13] Trend Six: Telecom Cloud Evolution - The telecom cloud is evolving from traditional cloud infrastructure to AI-native infrastructure, enabling operators to transition from "connectivity pipelines" to "connectivity + intelligent computing." This evolution is crucial for supporting the comprehensive intelligence of operator networks [14] Trend Seven: Agentic Core - The trend of agentification in AI terminals and services is driving the agentification of network architecture, enhancing differentiated connectivity capabilities and creating new business paradigms for on-demand network usage. The global digital services market based on agentified networks is projected to exceed $2.5 trillion by 2030 [15]

Core Viewpoint - The 2025 China International Information Communication Exhibition (PT Exhibition) showcased Nokia Bell's AI-driven self-intelligent network operation solutions, emphasizing the transition towards fully autonomous networks [2] Group 1: Self-Intelligent Network Development - The evolution of self-intelligent networks ranges from Level 0 (fully manual operation) to Level 5 (completely autonomous operation) [2] - Chinese operators are currently at approximately Level 3.5, with some scenarios achieving Level 4, which still requires human supervision [2] Group 2: Key Pathways to Full Autonomy - The four key pathways to achieving fully autonomous networks include: 1. Trustworthy AI-Ready data, ensuring data quality and organization for AI [2] 2. Measurable fully automated processes, transforming operational models and eliminating process bottlenecks [2] 3. Evolving intelligent agents capable of reasoning, acting, and self-learning for accurate intent management [2] 4. Predictive "network twins" that provide real-time reflections of network status through holographic topology [2]

Financial Performance - In the first half of 2025, the company achieved operating revenue of 106.04 million yuan, a decrease of 19.16% compared to the same period last year [2][3] - The net profit attributable to shareholders was -43.04 million yuan, down 85.70% year-on-year [3] - The net profit after deducting non-recurring gains and losses was -44.06 million yuan, a decline of 71.40% compared to the previous year [3] Business Challenges - The decline in software development orders due to telecom operators' cost-cutting strategies negatively impacted revenue targets for the first half of 2025 [3] - The company is actively advancing the development of the next-generation OSS product platform to create more business opportunities [3] Project Updates - The Zhengzhou Airport computing power project is progressing steadily, with equipment procurement and network testing underway [4] - The project is expected to generate new business income over a five-year operational period once completed [4] Strategic Direction - The company maintains a clear operational strategy, focusing on customer value and integrated marketing efforts [4] - New projects, including the Zhengzhou Airport Intelligent Computing Center, are being executed, with significant progress in product development [4] Capital Raising - The company has initiated a private placement for 2024, pending approval from the Shenzhen Stock Exchange and the China Securities Regulatory Commission [6]

Core Insights - The launch of the all-domain integrated computing network scheduling platform and the first wide-area network model and multi-agent system marks a significant advancement in China's computing resource management and network operations [1][2] - The platform aims to connect regions with abundant computing resources to those with high demand, potentially reducing management costs for enterprises by 50% [1] - The AI automatic design toolchain for baseband circuits is expected to streamline the design process, reduce costs, and improve efficiency in communication systems [2] Group 1 - The all-domain integrated computing network scheduling platform utilizes three pioneering technologies to achieve integrated scheduling across the nation [1] - The platform has been tested in various sectors, including hospitals and schools, and is aligned with national initiatives like "East Data West Computing" [1] - The wide-area network model and multi-agent system can autonomously sense, maintain, and repair network issues, applicable in 27 typical network operation scenarios [1][2] Group 2 - The system, led by the Purple Mountain Laboratory in collaboration with Jiangsu Future Network Group, China Unicom, and Huawei, has achieved an overall intelligence capability level of L4 [2] - The goal is to reach L5 level, enabling a fully self-intelligent network that can be generated on demand [2] - The AI automatic design toolchain can reduce the area of baseband circuits by 70% while maintaining communication performance, significantly shortening the design iteration process to just a few hours [2] Group 3 - The advancements in baseband circuit design are expected to lower energy consumption and costs for enterprises, ultimately benefiting consumers with improved product performance at reduced prices [3]

Performance Forecast - The company expects a net loss attributable to shareholders of 40 million to 46 million yuan for the current reporting period, compared to a loss of 23.1757 million yuan in the same period last year [1] - The net loss after deducting non-recurring gains and losses is projected to be between 41 million and 47 million yuan, compared to a loss of 25.7039 million yuan in the previous year [1] - The basic earnings per share are expected to be a loss of 0.39 to 0.45 yuan per share, compared to a loss of 0.23 yuan per share in the same period last year [1] Communication with Auditors - The company has communicated with the accounting firm regarding the performance forecast and there are no discrepancies between the company and the auditors concerning the performance forecast [1] Reasons for Performance Variance - The company's operating income shows significant seasonality, with a higher proportion in the fourth quarter, while management, sales, and R&D expenses are incurred more evenly throughout the quarters, leading to losses in the first half of 2025 [1] - The decline in software development business orders due to telecom operators' cost-reduction strategies has impacted the revenue targets for the first half of 2025 [1] - The company is actively advancing the development of the next-generation OSS product platform to create more business opportunities for incremental revenue in OSS [1] - In the new computing power business, the company is in a rapid development phase, with related projects progressing orderly towards operational status to gradually realize revenue [1]

Core Insights - Huawei's data center autonomous network achieved an overall intelligence level of AN L3.9, with core business scenarios in "planning and configuration" and "application deployment and configuration" reaching AN L4 level, marking it as the first solution to pass EANTC L4 autonomous network testing [1][2] - The evaluation by EANTC, a recognized independent network testing organization, serves as a significant reference for assessing network product performance, reliability, and intelligence level [2] Group 1 - The autonomous network grading assessment focused on three key value scenarios: planning and configuration, application deployment and issuance, and monitoring and troubleshooting, analyzing four critical capabilities of autonomous networks [2] - Huawei's solution demonstrated high intelligence and reliability in data center network operations, aligning well with the goals of AN L4, especially as enterprises expand data centers due to AI/ML workloads and private cloud adoption [2][3] Group 2 - The autonomous network solution has been deployed at scale in industries such as finance and the internet, with plans for broader implementation across various sectors in collaboration with industry organizations and testing agencies [3]

Core Viewpoint - The global telecommunications industry is facing a "growth without revenue" bottleneck, where the increase in network traffic does not translate into corresponding revenue growth, leading to a "growth reduction" phenomenon [1][3]. Group 1: Industry Challenges - The traditional network service model is homogenized, failing to meet diverse market demands, resulting in a lack of differentiation between high-value and ordinary services [1]. - Operators are experiencing a "pipeline" dilemma, where the expansion of 5G networks does not yield proportional income increases [1]. Group 2: Differentiated Connectivity - "Differentiated connectivity" has become a consensus in the industry, focusing on using technology to identify the value of different services and providing customized network guarantees for high-priority applications [3]. - Singapore Telecommunications (Singtel) exemplifies this approach by offering tiered service packages based on user needs, transforming operators from mere traffic providers to "service value definers" [3]. Group 3: Programmable Networks and AI Integration - Ericsson views programmable networks as a core development goal for 5G-A, enabling dynamic network configuration based on business needs through software-defined networking (SDN) and network function virtualization (NFV) [5]. - The integration of AI is reshaping network management, transitioning from passive to proactive management, with AI acting as a "network brain" for autonomous decision-making [5][7]. Group 4: Ecosystem Development - Ericsson is building an open ecosystem to attract third-party developers, creating a "network as a service" economic model, with a RAN ecosystem comprising 56 members [8]. - The acquisition of Vonage and the establishment of Aduna aim to create a CPaaS ecosystem, addressing fragmentation in network capability and enhancing collaboration with developers [8]. Group 5: Future Outlook - Despite the commercial launch of 5G-A, challenges remain, including deployment strategy delays in some regions and low penetration rates of new terminals [9]. - Ericsson proposes a dual-track development strategy to enhance 5G-A capabilities while promoting end-to-end differentiated services, aiming to redefine the telecommunications industry's revenue model and digital infrastructure [9].

Core Insights - Huawei has been awarded the "Data Center Intelligent Network Basic Capability Solution Testing Certificate," marking a significant milestone in the development of intelligent network technology [1][2] - The certification validates Huawei's technical capabilities in L4 intelligent networks for data centers, setting a new technical benchmark in the industry [2] Group 1: Certification and Evaluation - The China Academy of Information and Communications Technology (CAICT) issued the certification to Huawei and other companies, confirming their capabilities in intelligent networks [1] - CAICT, in collaboration with industry partners, developed the "Intelligent Network Basic Capability Grading Assessment Testing Method - Data Center Network (2023 Edition)," which defines L1 to L4 grading capabilities and testing methods across three major value scenarios [1] Group 2: Performance and Applications - Huawei's intelligent network solution excelled in three key scenarios: intelligent deployment of POD, intelligent application access initiation, and intelligent fault handling, demonstrating capabilities such as automatic generation of optimal deployment plans and precise fault prediction [1][2] - The solution enhances operational efficiency for industries like finance by providing self-generating configurations and self-diagnosing faults [2] Group 3: Future Developments - Huawei plans to continue increasing R&D investment and collaborate with industry partners to advance intelligent network technology, contributing to the construction of smart, efficient, and reliable ICT infrastructure for digital transformation [2]