在这一背景下，"超节点"作为突破传统算力瓶颈的新型产品形态，正逐步成为行业主流。华为副董事 长、轮值董事长徐直军在华为全联接大会2025上表示，从大型AI算力基础设施建设的技术方向看，超 节点已经成为主导性产品形态，并正在成为AI基础设施建设的新常态。 在超节点赛道上，国内科技巨头已率先交出答卷。华为此前已推出CloudMatrix384超节点，将384颗昇 腾NPU和192颗鲲鹏CPU通过高速网络MatrixLink全对等互联，形成一台超级"AI服务器"，单卡推理吞吐 量达到2300 Tokens/s。近期，阿里云发布新一代磐久128超节点AI服务器，集成阿里自研CIPU 2.0芯片 和EIC/MOC高性能网卡，单柜支持128个AI计算芯片。 随着大模型参数进入爆炸式增长阶段，AI算力需求正加速从单点突破走向系统级整合。 从全球来看，人工智能基础设施建设正按下"加速键"，英伟达、OpenAI、Meta等科技巨头纷纷抛出AI 基建新计划，推动AI算力规模发展。华龙证券认为，中美AI竞争正从"单卡性能"走向"系统级效率"，中 国正在用集群建设+开源生态+工程化交付的方式完成AI基建方面的弯道超车。 国产阵营发力 ...

9月18日，在华为全联接大会2025期间，以"共建智能世界云底座，加速政务与产业智能升级"为主题的 AI城市峰会隆重举行。此次峰会聚焦城市"智"理、智慧政务与产业数智升级三大领域，汇集行业专家、 政府代表和生态伙伴，从政策趋势解读到AI CITY政务服务、城市治理、产业发展国内外实践经验分享 及国内典型案例发布，共同探讨人工智能技术快速演进趋势下，如何驱动"AI CITY"体系化创新，赋能 智慧城市建设步入数智化转型新阶段。 当前，该架构已在实践中初显价值：在政务服务领域，携手深圳市推动深小i智能化升级，70名AI数智 员工覆盖240个场景，日均服务超15万次；在城市治理领域，广州白云区构建"1+4+N"智慧城管体系， 推动多项城市管理场景"革命性重塑"；在产业赋能领域，广州花都区新型工业化平台，助力企业研发周 期从数月缩至数天，产品设计效率提升50%。东莞人工智能大模型中心，帮助制造企业将产线检测准确 率提至99%。 对于华为云AI CITY未来前景展望，曹既斌提出三点倡议：希望携手更多科研院所和政企客户，不断创 新打造城市政务与产业智能升级先锋案例；邀请各方伙伴加入AI CITY产业联合创新计划，在建设和 ...

Core Viewpoint - The AI City Summit held during Huawei's Connect 2025 focuses on the integration of artificial intelligence in urban governance and industrial upgrades, emphasizing the transition to a smart city framework driven by AI technology [1][3]. Group 1: AI CITY Framework - Huawei and the National Information Center released the "AI CITY 2025 Government and Industry Pioneer Cases," highlighting the transformative impact of AI on urban governance and development [3]. - The AI CITY "1234MNX" framework addresses challenges in urban governance and industrial upgrades, moving beyond traditional data-driven smart cities to a model driven by "models + knowledge" [5][6]. - The framework includes a unified intelligent cloud base, a dual security system, three AI engineering platforms, and various operational categories to enhance resource efficiency and innovation [6]. Group 2: Practical Applications and Achievements - The AI CITY framework has shown initial value in practical applications, such as the intelligent upgrade in Shenzhen, which serves over 150,000 users daily across 240 scenarios [6]. - In Guangzhou, the "1+4+N" smart city management system has revolutionized urban management, while the industrial platform in Huadu has significantly reduced product design cycles and improved efficiency [6]. - The Dongguan AI model center has increased manufacturing line inspection accuracy to 99%, showcasing the practical benefits of AI integration in various sectors [6]. Group 3: Future Prospects and Initiatives - Huawei's future vision for AI CITY includes collaboration with research institutions and enterprises to innovate urban governance and industrial upgrade cases [7]. - The company aims to enhance AI talent cultivation to support long-term industrial development and invites partners to join the AI CITY industry innovation plan [7]. - Huawei emphasizes the importance of a robust computing foundation and the development of AI models to support urban applications and ecosystem growth [8][12]. Group 4: AI's Role in Urban Development - AI is positioned as a core driver for transforming urban production, lifestyle, and governance, with significant implications for public services and industrial applications [11]. - Various successful AI applications have been implemented across cities, such as the data circulation platform in Shanghai and the digital assistant in Guizhou, enhancing service efficiency and user experience [11][12]. - Huawei's commitment to building an open and shared computing system aims to integrate global capabilities into local AI ecosystems, fostering innovation and collaboration [12].

明敏 发自 凹非寺 量子位 | 公众号 QbitAI 华为云算力再迎重大突破！ 刚刚落幕的华为全联接大会2025，一系列新进展发布—— 这距离CloudMatrix384超节点2025年4月正式发布仅半年，期间其 能力持续进化 ： 现阶段， AI行业内依旧被算力焦虑笼罩 。硅谷大厂近期在算力、芯片领域动作频频： OpenAI一边和博通自研AI芯片，一边向甲骨文抛出3000亿美元买算力；马斯克百天建成万卡超算集群，还计划向百万卡规模冲击，同时悄悄 布局芯片；Meta、AWS等企业也在积极获取更多算力资源……但算力的发展并非一蹴而就，它需要在单点技术上极致突破，还涉及芯片、硬 件、架构、软件、网络、能源乃至整个产业生态的协同演进。 放眼全球，能够输出澎湃算力的供应商，都离不开十数年、数十年的沉淀积累。 华为云作为其中一员，探索路径因所处产业阶段而显得尤为深刻：不仅需要在技术"无人区"重新定义算力运行规则；还需把握AI发展时机，通 过快速迭代响应产业海量需求。一步步成长为今天的"算力黑土地"。 AI算力云服务升级， 基于华为云刚刚发布的最新AI服务器规划， CloudMatrix的云上超节点规格将从384卡升级到未 ...

人民财讯9月18日电，9月18日，记者从华为全联接大会2025上了解到，华为云Stack8.6新版本适配 CloudMatrix384超节点，将于10月30日正式商用。这也是业界首个支持CloudMatrix384超节点的混合 云。 ...

人民财讯9月18日电，9月18日，华为全联接大会2025期间，华为混合云总裁肖霏正式发布华为云 Stack8.6，并宣布华为云Stack8.6已正式适配CloudMatrix384超节点。目前，全球超过5500家政企客户选 择华为云Stack开展云上创新。 ...

Group 1 - Huawei's Full Connection Conference 2025 will be held from September 18 to 20, focusing on "Elevating Industry Intelligence" and showcasing Huawei's comprehensive intelligent strategy and new digital infrastructure products [1] - The conference will feature the Ascend AI Industry Summit, where the Ascend scenario-based integrated machine will be launched [1] - Huawei Cloud has introduced the CloudMatrix384 super node, which integrates 384 Ascend NPUs and 192 Kunpeng CPUs, achieving a computing power of 300 PFlops [1] Group 2 - Advanced Communication is a diamond distributor for Huawei, responsible for the sales of Huawei's Ascend integrated machines [2] - Softcom Power has collaborated with a major telecommunications equipment company to launch the AI robot - Ascend integrated machine [3]

Core Viewpoint - The importance of chips is acknowledged, but the ability to provide the required computational results for customers is emphasized as more critical [1] Group 1: Huawei Cloud's Developments - Huawei Cloud is undergoing significant organizational restructuring, with a focus on enhancing computational power through its Ascend AI cloud services and Tokens services [1] - The CloudMatrix384 super node, launched in April, integrates 384 Ascend NPUs and 192 Kunpeng CPUs, achieving a computational scale of 300 PFlops [2] - The Tokens service has been integrated with the CloudMatrix384 super node, achieving a maximum of 2400 TPS and 50 ms latency, surpassing industry standards [2][3] Group 2: Performance and Growth Metrics - Huawei Cloud's overall computational capacity has increased by nearly 250% compared to the previous year, with the number of Ascend AI cloud service customers rising from 321 to 1714 [5] - The CloudMatrix384 super node can support training of large models, with the capability to connect 432 super nodes to form a 160,000-card AI cluster [2] - The deployment of over 40 CloudMatrix384 super nodes in Guizhou is part of a strategy to create a national computational network [5] Group 3: Market Position and Strategic Focus - Huawei Cloud ranks second in China's cloud service market with an 18% share, while Alibaba Cloud holds a 33% share [6] - The market demand is shifting from "cloud adoption" to "AI integration," prompting Huawei Cloud to streamline its operations and focus on maximizing the advantages of its Ascend AI and Pangu model combinations [6][7] - The organizational restructuring aims to concentrate resources on core areas that can leverage AI capabilities effectively [6][7]

Financial Performance - Huawei achieved revenue of 427.04 billion yuan in the first half of the year, a year-on-year increase of 3.95%, while net profit decreased by 32% to 37.195 billion yuan [1] - R&D investment reached 96.95 billion yuan, up 9.04% year-on-year, accounting for 22.7% of total revenue [1] - Despite revenue growth, Huawei's profitability is under pressure due to a complex competitive environment [1] Market Position - Huawei regained the top position in the Chinese smartphone market after four years, with a shipment of 12.5 million units, despite a 4.1% decline in the overall market [1] - The company's consumer business, smart automotive, and Huawei Cloud are identified as key performance drivers [1] Automotive Sector - Huawei's automotive division, Hongmeng Zhixing, aims to deliver over 900,000 vehicles by August 2025, with a projected revenue growth of 4.7 times in 2024 and achieving profitability in its first year [1] - The automotive sector faces increasing competition from companies like Li Auto and Xpeng, which are launching high-cost-performance models [2] Cloud Computing and AI - Huawei is focusing on cloud computing and AI as critical future growth areas, building an ecosystem around Ascend processors and MindSpore framework [2] - The launch of CloudMatrix384 super node, featuring 384 Ascend NPUs and 192 Kunpeng CPUs, aims to enhance AI computing capabilities, achieving a performance scale of 300 PFlops [2] - Huawei Cloud's overall computing power has increased by nearly 250% year-on-year, with the number of clients using Ascend AI cloud services rising from 321 to 1,714 [3]

Core Insights - The demand for computing power is significantly increasing due to the rise of large models, with Huawei Cloud aiming to build a "national computing network" to meet this demand [1][4] - Huawei Cloud's overall computing power has grown by nearly 250% compared to the same period last year, with the number of clients using Ascend AI cloud services increasing from 321 to 1714 [1][2] - Huawei Cloud's CloudMatrix384 super node can achieve a computing power of 300 PFlops and can support training of large models with up to 130 billion parameters [2] Group 1: Computing Power Demand - Computing power is identified as the core infrastructure of the intelligent world, with a projected growth of tens of thousands of times in demand over the next decade [1] - Huawei Cloud is focusing on creating a "computing black land" around key hubs to enhance computing capabilities [1] Group 2: Technological Innovations - The CloudMatrix384 super node integrates 384 Ascend NPUs and 192 Kunpeng CPUs, forming a super AI server capable of extensive model training [2] - Huawei Cloud's Tokens service has performance advantages in high-throughput scenarios, achieving a new industry high of 2400 TPS with a latency of 50ms [2] Group 3: Industry Applications - Huawei Cloud's computing services are widely used across various sectors, including central state-owned enterprises, intelligent driving, large models, and consumer electronics [2][3] - The CloudVeo intelligent driving cloud service has been launched, establishing a dual-zone layout in the automotive sector [3] Group 4: Strategic Focus - Huawei Cloud is accelerating the construction of computing power fields nationwide to seize strategic advantages in the wave of large models and intelligence [4] - Recent organizational restructuring aims to enhance focus on AI development and allocate more resources to the AI and computing power industries [4]