华为正通过从大模型设计到基础设施的软硬协同，探索构建其全栈 AI 竞争力的路径。 华为 AI 发展策略已逐渐从追赶并对标业界 SOTA 模型，转向为更好地 发挥自研昇腾硬件潜力而量身定做模型架构。这一双向协同进化路径，旨在解决 AI 模型规模化应用中的系统性问题，并构建一个由软硬件协同架构、算子与 软件栈构成的全栈技术体系。 盘古大模型的演进，其核心是为解决大规模分布式系统中的效率难题。 随着大语言模型从稠密架构全面转向混合专家（ MoE ）稀疏架构，业界普遍面临专家 负载不均衡这一系统性瓶颈，它制约了 MoE 模型在训练和推理中的实际性能。华为将此系统性问题作为其软硬架构创新的核心方向，标志着其关注点已从单 纯硬件或单纯 AI 算法问题，拓展至在自研硬件上更高效解决 AI 系统工程问题。 华为在大模型层面并行推出了两种创新路径。一方面， Pangu Pro MoE 通过架构破局，提出分组专家混合（ MoGE ）架构，旨在通过结构性设计解决负载不 均衡问题。另一方面 ， Pangu Ultra MoE 则通过系统级优化，以仿真先行的设计方法来优化模型架构从而更好的适配昇腾硬件，并通过贯穿训练和推理的协 同优化 ...

Investment Rating - The report does not explicitly state an investment rating for the industry. Core Insights - Huawei is exploring a "soft and hard integration" strategy to enhance its AI competitiveness, transitioning from merely catching up with industry SOTA models to customizing model architectures for its self-developed Ascend hardware [12][30]. - The evolution of the Pangu model series reflects a shift from parameter competition to a focus on efficiency and scalability, culminating in the adoption of the Mixture of Experts (MoE) architecture [12][30]. - The report highlights the introduction of innovative architectures like Pangu Pro MoE and Pangu Ultra MoE, which aim to maximize the utilization of Ascend hardware through structural and system-level optimizations [36][62]. Summary by Sections 1. Evolution of Pangu Models - The Pangu model series began with PanGu-α, a 200 billion parameter model, which established a technical route based on Ascend hardware [12][30]. - PanGu-Σ, launched in 2023, marked an early attempt at sparsification, exploring trillion-parameter models with a focus on efficiency [15][18]. - Pangu 3.0 introduced a "5+N+X" architecture aimed at deep industry applications, showcasing its capabilities in various sectors [22][23]. 2. Pangu Pro MoE and Pangu Ultra MoE - Pangu Pro MoE addresses the challenge of expert load imbalance in distributed systems through a new architecture called Mixture of Grouped Experts (MoGE) [36][37]. - The MoGE architecture ensures load balancing by structuring the selection of experts, thus enhancing efficiency in distributed deployments [45][46]. - Pangu Ultra MoE emphasizes system-level optimization strategies to explore the synergy between software and hardware, reflecting a practical application of the soft and hard integration concept [62]. 3. CloudMatrix Infrastructure - CloudMatrix serves as the physical foundation for AI infrastructure, enabling high-performance communication and memory management across distributed systems [5][10]. - The infrastructure supports the Pangu models by providing a unified addressing distributed memory pool, which reduces performance discrepancies in cross-node communication [5][10]. 4. Full-Stack Collaboration - Huawei's AI strategy is centered around full-stack collaboration, integrating open-source strategies to build an ecosystem around Ascend hardware [10][12]. - The architecture, systems, and operators form the three pillars of this full-stack collaboration, aimed at enhancing the overall efficiency and effectiveness of AI solutions [10][12].

Core Viewpoint - Huawei has officially announced the open-sourcing of its Pangu models, including the 70 billion parameter dense model and the 720 billion parameter mixture of experts (MoE) model, as part of its Ascend ecosystem strategy to promote AI technology across various industries [1][2]. Group 1: Model Development and Open-Sourcing - Huawei has launched the Pangu Pro MoE 72B model weights and basic inference code on an open-source platform, with plans to release the Pangu 7B model weights and inference code soon [1]. - The Pangu Pro MoE model, with 720 billion parameters and 160 billion active parameters, has demonstrated performance comparable to larger models, ranking first among domestic models with fewer than 1 trillion parameters on the SuperCLUE leaderboard [1]. - The company plans to open-source the Pangu 72B MoE model first, followed by smaller models potentially for academic institutions [2]. Group 2: Technical Advancements - Huawei has introduced a new model, the Pangu Ultra MoE, with a parameter scale of 718 billion, trained entirely on the Ascend AI computing platform [2]. - The model training efficiency has been highlighted, with a model compute utilization (MFU) of 41% achieved in pre-training and over 50% in specific configurations [3]. - The architecture of the Ascend super nodes has been optimized for extreme parallelism, enhancing training efficiency and inference performance [3]. Group 3: Ecosystem and Future Plans - Huawei is committed to improving its ecosystem and ensuring compatibility with mainstream industry ecosystems to support customer development [2]. - The company has also announced upgrades to its Pangu models for natural language processing, computer vision, multimodal applications, prediction, and scientific computing at the Huawei Developer Conference [3].

Core Insights - The article highlights significant advancements in China's artificial intelligence (AI) sector, particularly with the launch of Huawei's Pangu Ultra MoE model, which has a parameter scale of 718 billion, showcasing the capability of domestic computing power to train world-class large models [1][2] - The competition between China and the United States in AI is characterized by a "strong U.S. and fast China" dynamic, where China is rapidly closing the gap through application innovation, data scale, and policy support [1][2] - China's AI industry has made notable progress, becoming the largest holder of AI patents globally, with a core industry scale nearing 600 billion yuan and over 4,700 companies, indicating a comprehensive industrial system [3][4] Industry Analysis - Computing power is identified as a critical battleground in AI development, with talent, data, and computing power being the three key elements [2] - Despite the existing gap in core algorithms and advanced computing power, China is leveraging innovative approaches to enhance system performance, demonstrating a pathway to overcome technological barriers [2][3] - The article emphasizes the importance of a systematic approach to AI development, highlighting China's full-stack autonomous technology chain that is narrowing the gap with global leaders [3] Strategic Outlook - The development of AI in China requires confidence and patience, as it involves a comprehensive competition of innovation systems, industrial resilience, and strategic vision [4] - China's manufacturing sector, which accounts for approximately 30% of global manufacturing value added, serves as a significant advantage for AI development [4] - Continuous improvement in high-end chip architecture, cluster communication efficiency, and software ecosystems is essential for the advancement of China's AI industry [3][4]

Core Insights - Huawei has made significant advancements in the MoE model training field by launching a new model called Pangu Ultra MoE, which has a parameter scale of 718 billion [1] - The model is trained on the Ascend AI computing platform and represents a near-trillion MoE model, showcasing the performance leap in ultra-large-scale MoE training [1] - Huawei has released a technical report detailing the architecture and training methods of the Pangu Ultra MoE model, highlighting innovative designs to address challenges in training stability for ultra-large-scale and highly sparse MoE models [1]

Core Insights - Shanghai's AI industry cluster is evolving, integrating traditional industries with modern services through advanced computing power and algorithms [1][8] - The shift from traditional methods to AI-driven processes is enhancing efficiency and quality in sectors like steel manufacturing and insurance [2][4] Group 1: Steel Industry Innovations - Baosteel is utilizing AI for predictive furnace condition monitoring, achieving over 90% accuracy in temperature predictions and 96% accuracy in surface defect identification [2][3] - The implementation of AI applications is estimated to generate over 10 million yuan in direct economic benefits annually for Baosteel [2] - Baosteel plans to launch 300 AI application scenarios by 2025, establishing five benchmark smart production lines [3] Group 2: Insurance Sector Transformation - China Pacific Insurance is developing a proprietary large model infrastructure, improving training efficiency by 30% and enhancing claims review accuracy by 59.4% [4][5] - AI technologies are being fully integrated into insurance operations, leading to an 80.5% customer satisfaction rate [4] - The company aims to promote international strategies and establish a carbon emission monitoring system in collaboration with leading firms [5] Group 3: AI Infrastructure Development - Shanghai Supercomputing Center is creating a public AI computing service platform, becoming a central hub for AI innovation in the Yangtze River Delta [6][7] - The platform is designed to optimize resource allocation among over 80 participating enterprises, enhancing the efficiency of AI model training [6] - The Shanghai government aims to establish a world-class AI industry ecosystem by 2025, targeting a computing power scale exceeding 100 EFLOPS [8]

Core Insights - The Huawei AI + Manufacturing Industry Summit 2025 was successfully held in Guangzhou, focusing on the integration of AI technologies into the manufacturing sector to reduce costs and enhance efficiency [1] - AI is seen as a significant driver for cost reduction and innovation in the manufacturing industry, creating new survival rules for enterprises [2][4] Group 1: AI's Role in Manufacturing - AI technology is penetrating various stages of manufacturing, including R&D, production, and supply chain management, leading to improved market insights, optimized product design, and enhanced production efficiency [4] - The integration of AI with IoT and cloud computing is automating and smartening production processes, significantly increasing factory agility and productivity [4][6] Group 2: Infrastructure and Data Management - The construction of new infrastructure is essential for the digital transformation of the manufacturing industry, facilitating data flow and resource optimization [6][7] - Huawei emphasizes breaking down "data silos" through unified standards, which is crucial for data-driven production decisions [7] Group 3: Collaborative Efforts and Solutions - Huawei collaborates with industry partners to develop comprehensive solutions that address the diverse AI computing needs across various sectors, including automotive, semiconductor, and biopharmaceuticals [8] - The company has created over 20 solutions across seven major scenarios, aiming to enhance the digital transformation capabilities of enterprises [8] Group 4: Future Vision - Huawei aims to accelerate the intelligent transformation of the manufacturing industry, positioning itself as a key player in driving innovation and development [9]