Core Viewpoint - Huawei demonstrates strong technological capabilities in the semiconductor industry, particularly with its Ascend series chips and the recent launch of CM384, positioning itself as a leader in domestic AI chips [2][3]. Group 1: Financial Performance - In 2024, Huawei achieved a total revenue of RMB 862.072 billion, representing a year-on-year growth of 22.4% [5]. - The smart automotive solutions segment saw a remarkable revenue increase of 474.4%, while terminal business and digital energy businesses grew by 38.3% and 24.4%, respectively [5]. - Revenue from the Chinese market reached RMB 615.264 billion, driven by digitalization, intelligence, and low-carbon transformation [5]. Group 2: Huawei Cloud - The overall public cloud market in China is projected to reach USD 24.11 billion in the second half of 2024, with IaaS accounting for USD 13.21 billion, representing a year-on-year growth of 14.4% [6]. - Huawei Cloud holds a 13.2% market share in the Chinese IaaS market, making it the second-largest cloud provider after Alibaba Cloud [6]. - Huawei Cloud's revenue growth rate reached 24.4%, the highest among major cloud vendors in China [6]. Group 3: Ascend Chips - The CloudMatrix 384 super node integrates 384 Ascend 910 chips, achieving a cluster performance of 300 PFLOPS, which is 1.7 times that of Nvidia's GB200 NVL72 [10]. - The single-chip performance of Huawei's Ascend 910C is approximately 780 TFLOPS, which is one-third of Nvidia's GB200 [10][11]. - The Ascend computing system encompasses a comprehensive ecosystem from hardware to software, aiming to meet various AI computing needs [15][20]. Group 4: HarmonyOS - HarmonyOS features a self-developed microkernel, AI-native capabilities, distributed collaboration, and privacy protection, distinguishing it from Android and iOS [12]. - The microkernel architecture enhances performance and fluidity, while the distributed soft bus technology allows seamless connectivity among devices [12][13]. Group 5: Kirin Chips - The Kirin 9020 chip has reached high-end processor standards, comparable to a downclocked Snapdragon 8 Gen 2 [23]. - The Kirin X90 chip, based on the ARMv9 instruction set, features a 16-core design with a frequency exceeding 4.2GHz, achieving a 40% improvement in energy efficiency [25][26]. Group 6: Kunpeng Chips - Kunpeng processors are designed for servers and data centers, focusing on high performance, low power consumption, and scalability [27]. - The Kunpeng ecosystem strategy emphasizes hardware openness, software open-source, enabling partners, and talent development [29].

Summary of Huawei's CloudMatrix 8 Conference Call Company and Industry - **Company**: Huawei - **Industry**: Semiconductor and AI Computing Key Points and Arguments Product Overview - Huawei introduced the **CloudMatrix 8**, a powerful domestic solution in China built using the **Ascend 10C** chip, competing directly with Nvidia's **GB200 NVL72** [3][4] - The CloudMatrix 8 architecture is noted for its engineering advantages at the system level, not just at the chip level, with innovations across accelerator, networking, optics, and software layers [4] Performance Metrics - The CloudMatrix 8 can deliver **300 PFLOPS** of dense BF16 compute, nearly double that of the **GB200 NVL72** [10] - Key specifications comparison: - **BF16 dense PFLOPS**: CloudMatrix 300 vs. GB200 180 - **HBM capacity**: CloudMatrix 49.2 TB vs. GB200 13.8 TB - **HBM bandwidth**: CloudMatrix 1.229 TB/s vs. GB200 576 TB/s - **All-in System Power**: CloudMatrix 559,378 W vs. GB200 145,000 W [10][53] Power Consumption and Efficiency - The CloudMatrix 8 consumes significantly more power, drawing approximately **500 kW**, which is over **3.9 times** that of the GB200 NVL72 [51] - Despite higher power consumption, Huawei's system is designed to leverage China's abundant energy resources, allowing for scaling without power constraints [13][54] Supply Chain and Production Challenges - Huawei's Ascend chips are primarily produced by TSMC, with significant reliance on foreign production for components like HBM and wafers [16][19] - The company has reportedly circumvented sanctions to acquire necessary components, including **$500 million** worth of 7nm wafers [17] - Domestic production capabilities are improving, with SMC ramping up capacity, but foreign reliance remains a critical issue [24][27] Strategic Implications - The advancements in Huawei's technology are seen as a response to U.S. export controls, highlighting the importance of AI competitiveness as a national security concern [9] - The CloudMatrix 8's design reflects a strategic focus on scaling up capabilities, leveraging domestic strengths in networking and infrastructure software [11][15] Market Positioning - Huawei's CloudMatrix 8 is positioned as a competitive alternative to Nvidia's offerings, with a focus on system-level performance rather than just chip performance [5][6] - The architecture's design allows for significant scaling, which is crucial for meeting the demands of AI workloads [28][30] Conclusion - Huawei's CloudMatrix 8 represents a significant advancement in China's AI computing capabilities, with a focus on system-level innovations and leveraging domestic resources, despite challenges in supply chain and power efficiency [54]

Core Viewpoint - Nvidia has been authorized to resume exports of its H20 GPU to China, but Huawei's CloudMatrix 384 system, showcased at the World Artificial Intelligence Conference, presents a formidable alternative with superior specifications [3][4]. Summary by Sections Nvidia H20 GPU and Huawei's CloudMatrix 384 - Nvidia's H20 GPU may have sufficient supply, but operators in China now have stronger alternatives, particularly Huawei's CloudMatrix 384 system, which features the Ascend P910C NPU [3]. - The Ascend P910C promises over twice the floating-point performance of the H20 and has a larger memory capacity, despite being slower [3][6]. Technical Specifications of Ascend P910C - Each Ascend P910C accelerator is equipped with two computing chips, achieving a combined performance of 752 teraFLOPS for dense FP16/BF16 tasks, supported by 128GB of high-bandwidth memory [4]. - The CloudMatrix 384 system is significantly larger than Nvidia's systems, with the ability to scale up to 384 NPUs, compared to Nvidia's maximum of 72 GPUs [11][9]. Performance Comparison - In terms of memory bandwidth and floating-point performance, the Ascend P910C outperforms Nvidia's H20, with 128GB of HBM compared to H20's 96GB [6]. - Huawei's CloudMatrix system can support up to 165,000 NPUs in a training cluster, showcasing its scalability [11]. Inference Performance - Huawei's CloudMatrix-Infer platform enhances inference throughput, allowing each NPU to process 6,688 input tokens per second, outperforming Nvidia's H800 in terms of efficiency [14]. - The architecture allows for high-bandwidth, unified access to cached data, improving task scheduling and cache efficiency [13]. Power, Density, and Cost - The estimated total power consumption of the CloudMatrix 384 system is around 600 kW, significantly higher than Nvidia's NVL72 at approximately 120 kW [15]. - The cost of Huawei's CloudMatrix 384 is around $8.2 million, while Nvidia's NVL72 is estimated at $3.5 million, raising questions about deployment and operational costs [16]. Market Dynamics - Nvidia has reportedly ordered an additional 300,000 H20 chips from TSMC to meet strong demand from Chinese customers, indicating ongoing competition in the AI accelerator market [17].

Core Viewpoint - Huang Renxun, CEO of NVIDIA, is actively engaging with the Chinese market despite ongoing U.S. sanctions on semiconductor exports to China, highlighting the importance of China as a critical market for NVIDIA's growth and future opportunities [5][12][16]. Group 1: NVIDIA's Market Position and Challenges - NVIDIA has achieved a market capitalization exceeding 4 trillion yuan, driven by the global AI boom, but faces significant challenges due to U.S. export restrictions on its A100 and H100 chips to China [4][23]. - The company’s revenue from the Chinese market reached $17.1 billion in 2024, marking a 66% year-on-year increase, contributing 13% to NVIDIA's total revenue [17][18]. - The U.S. government's strict AI chip export regulations have led to a significant decline in NVIDIA's market share in Asia, dropping from 95% to 50% [20]. Group 2: Huang Renxun's Engagement with China - Huang Renxun has made multiple visits to China, emphasizing the importance of the Chinese market and expressing a desire to continue collaboration with Chinese companies [15][16]. - During his visits, he has praised China's rapid AI development and robust supply chain, indicating a strong commitment to maintaining NVIDIA's presence in the market [15][17]. - Huang's efforts include addressing employee morale in China amidst fears of layoffs due to the impact of U.S. sanctions [6][14]. Group 3: Product Adaptations and Future Prospects - In response to export restrictions, NVIDIA has developed a "special supply version" of its H100 chip, named H20, which has significantly reduced performance but is tailored for the current needs of Chinese companies [25][26]. - Huang Renxun anticipates that the H20 chip will find success in the Chinese market, despite its limitations, as companies are eager to invest in AI capabilities [26]. - The emergence of domestic competitors in China, such as Huawei, poses a potential threat to NVIDIA's market dominance, especially as these companies advance their own chip technologies [27][28].

Investment Rating - The report maintains an "Outperform" rating for the computer industry [1][4]. Core Insights - The report highlights significant advancements in AI, particularly with the release of xAI's Grok 4, which claims to be the world's strongest AI model, achieving a training volume 100 times greater than its predecessor Grok 2 [3][60]. - Domestic AI chip companies are entering a concentrated IPO phase, with notable advancements in the domestic AI ecosystem, including the launch of Huawei's Ascend AI Cloud based on CloudMatrix 384 [3][4]. - The easing of EDA and H20 bans by the US is expected to accelerate the differentiation within the server industry, while the domestic market continues to push for localization and replacement of foreign systems [3][4]. Summary by Sections Industry Data - From January to May 2025, the software industry revenue reached 5.58 trillion yuan, growing by 11.2% year-on-year, with profit totaling 672.1 billion yuan, a 12.8% increase [13][14]. - The IC design sector showed the highest growth at 15.2%, while cloud and big data services also experienced significant growth [19][20]. AI Developments - The AI sector is witnessing rapid advancements, with Grok 4 outperforming other leading models in various assessments [60][65]. - Meta has initiated a talent acquisition strategy to bolster its AI capabilities, following setbacks with its Llama 4 model [74]. Localization Trends - The report notes that the domestic AI chip industry is moving towards a concentrated IPO phase, with significant developments in the domestic software ecosystem, including the number of applications for Huawei's Harmony OS surpassing 2,500 [3][4][19]. - The easing of US technology bans is seen as a catalyst for further domestic advancements and localization efforts [3][4]. Computing Power - The upcoming release of NVIDIA's next-generation AI server chip, GB300, is anticipated to enhance capital expenditures and chip procurement among major tech firms [4]. - The Ministry of Industry and Information Technology has launched initiatives to promote the construction of a unified national computing power service market [4].

Investment Rating - The report maintains a positive outlook on the computer industry, particularly focusing on the developments in stablecoins and supernodes [5][6]. Core Insights - The report emphasizes that stablecoins are transitioning from speculative concepts to mainstream payment solutions, supported by regulatory developments and strategic initiatives in regions like Shanghai and Hong Kong [7][8]. - The supernode architecture is highlighted as a significant advancement in computational power, with products like Huawei's CloudMatrix 384 demonstrating superior performance compared to traditional single-card systems [5][33]. Summary by Sections Stablecoins - The Shanghai Municipal State-owned Assets Supervision and Administration Commission recognizes the role of stablecoins in enhancing the financial system and encourages the exploration of related technologies [7][8]. - The report outlines that stablecoins are not merely a short-term trend but represent a long-term investment theme, with a shift towards legislative compliance and mainstream payment applications [6][9]. - Key companies involved in stablecoin technology and infrastructure include 恒生电子 (Hengsheng Electronics), 金证股份 (Jinzhen Co.), and others [5][6]. Supernodes - The report discusses the transition from single-card systems to supernode architectures, which integrate multiple GPUs for enhanced computational efficiency [33][34]. - Huawei's CloudMatrix 384 is presented as a benchmark for domestic supernode solutions, achieving 1.7 times the computational power and 3.6 times the memory capacity compared to Nvidia's GB200 NVL72 [55][56]. - Companies positioned to benefit from the supernode trend include 海光信息 (Haiguang Information), 浪潮信息 (Inspur Information), and 神州数码 (Digital China) [5][33].

Core Insights - The summit focused on the theme "Huawei Cloud Stack, Understanding Government and Enterprise in the Intelligent Era," highlighting the importance of AI technology in driving digital transformation for government and enterprise clients [1] - Huawei Cloud aims to enhance its AI cloud service products and solutions, collaborating with clients to implement AI technology across various business scenarios [1][3] - The company recognizes the diverse needs of government and enterprise users, categorizing them into four distinct roles to better tailor its offerings [4] Group 1: AI and Digital Transformation - Government and enterprise clients are increasingly adopting AI technologies for applications such as smart customer service, process optimization, safety supervision, and digital marketing [1] - Huawei Cloud's mixed cloud solutions are designed to support the digital transformation needs of these clients, providing over 120 cloud services and more than 50 scenario-based solutions [3] - The company emphasizes the importance of leveraging vast amounts of industrial, financial, and public data in China to foster AI industry growth [3] Group 2: User-Centric Approach - Huawei Cloud Stack has developed a comprehensive approach to support different user roles, including data center engineers, data engineers, AI algorithm model application engineers, and application development engineers [4] - The company is set to launch a mixed cloud solution compatible with CloudMatrix 384 super nodes, enabling clients to have their own cloud super nodes locally [4] Group 3: Case Studies and Applications - Xiangtan Steel Group, in collaboration with Huawei, launched the world's first steel industry large model, establishing a unified AI training center to enhance operational efficiency [5] - Chengdu Urban Investment Digital Group is building a trusted data space to empower the city's digital transformation, focusing on data integration and new business models [6] Group 4: Future Directions and Innovations - Huawei Cloud Stack is enhancing its capabilities in large model mixed cloud solutions, addressing challenges in computing power management and AI toolchain development [7] - The release of the white paper on "Government and Enterprise AI Platform Architecture and Application Practices" aims to provide a reference framework for clients to plan their AI platform construction effectively [7]

Core Insights - The report highlights a moderate recovery in the Chinese economy, with consumption and investment as core drivers, and suggests that the A-share market is suitable for medium to long-term investment due to its current valuation levels [8][9][11]. Domestic Market Performance - The Shanghai Composite Index closed at 3,387.40, with a slight decline of 0.04%, while the Shenzhen Component Index fell by 0.12% to 10,151.43 [3]. - The average price-to-earnings ratios for the Shanghai Composite and ChiNext are 13.90 and 37.06, respectively, indicating a mid-level valuation compared to the past three years [8][9]. International Market Performance - Major international indices such as the Dow Jones and S&P 500 experienced declines of 0.67% and 0.45%, respectively, while the Nikkei 225 rose by 0.62% [4]. Economic Policies and Developments - The State-owned Assets Supervision and Administration Commission (SASAC) reported that over 80% of key reform tasks for central and local state-owned enterprises have been completed as of Q1 2025 [5][8]. - The Henan provincial government has introduced guidelines to enhance high-quality investment attraction, promoting a dual approach of government and market-driven initiatives [5][8]. Industry Analysis - The software industry saw a revenue increase of 10.8% year-on-year in the first four months of 2025, with a notable rise in domestic AI chip localization from 20% to 34% [13]. - The lithium battery sector is projected to see significant growth, with a 43.97% year-on-year increase in new energy vehicle sales in China for the first five months of 2025 [17][25]. - The telecommunications sector reported a 1.0% year-on-year increase in revenue for the first four months of 2025, with a notable rise in 5G mobile phone users [26][27]. Investment Recommendations - The report suggests focusing on sectors such as consumer electronics, battery technology, and telecommunications for short-term investment opportunities, given their current performance and growth potential [8][9][11][29]. - In the lithium battery sector, attention is drawn to companies with strong research and development capabilities and those benefiting from the ongoing demand for electric vehicles [17][25]. Key Data Updates - The semiconductor industry is experiencing a recovery, with global sales expected to grow by 11.2% in 2025, driven by increasing demand for AI and consumer electronics [21][22][23]. - The chemical industry is seeing a slowdown in price declines, particularly in potassium and phosphorus fertilizers, indicating potential investment opportunities in these areas [19][20]. Sector Performance - The food and beverage sector showed resilience, with a 0.25% increase in May, despite challenges in the liquor segment, indicating a shift in consumer preferences towards other beverage categories [31][32]. - The electrical equipment sector is expected to benefit from increased domestic demand as the power grid construction accelerates, despite facing challenges from external trade policies [34][36]. Conclusion - The report emphasizes a cautiously optimistic outlook for various sectors, suggesting that investors should remain vigilant about policy changes and market dynamics while exploring opportunities in high-growth industries [8][9][11].

Core Viewpoint - Huawei has achieved significant advancements in training large models through its "Ascend + Pangu Ultra MoE" combination, enabling a fully controllable training process without the need for GPUs, showcasing industry-leading performance in cluster training systems [2][3]. Group 1: Technical Innovations - Huawei's training system has improved the model training efficiency significantly, with a pre-training model utilization rate (MFU) reaching 41% and a post-training throughput of 35K Tokens/s on the CloudMatrix 384 super node [3][34]. - The company has introduced a series of innovative solutions to address challenges in the MoE pre-training and reinforcement learning (RL) post-training processes, including intelligent parallel strategy selection and global dynamic load balancing [11][17]. - The training system utilizes a hierarchical All-to-All communication architecture to reduce communication overhead to nearly zero, enhancing the efficiency of expert parallel communication [14][15]. Group 2: Training Process Optimization - The training cluster's utilization has been optimized through a simulation-driven intelligent parallel optimization framework, which automates the selection of optimal deployment configurations [12][13]. - The team has implemented a memory optimization framework that achieves over 70% savings in activation memory, ensuring reliable long-term training even under increased memory pressure [25]. - The RL Fusion technology allows for flexible deployment modes, significantly improving resource scheduling during the inference phase and doubling the utilization rate in RL post-training [27][28]. Group 3: Model Specifications - The Pangu Ultra MoE model features 718 billion parameters, with a structure that includes 61 layers of Transformer architecture, designed for high sparsity and performance [32]. - The model's training utilized a cluster of 6K - 10K Ascend 800T A2 cards, achieving a high model utilization rate during the pre-training phase [32]. - The architecture supports efficient scaling to larger parameter models and clusters, with expectations of achieving an MFU greater than 50% in future iterations [32].

Core Viewpoint - The article discusses the rapid advancements in large language models (LLMs) and the challenges they face in inference, particularly regarding speed and energy efficiency. It highlights Huawei's innovative solutions to optimize these models through hardware-software integration, focusing on three key technologies that enhance inference speed and energy efficiency [2][4][11]. Group 1: Key Technologies - AMLA technology transforms complex multiplication into addition operations, significantly increasing chip utilization rates to 71% and improving performance by over 30% in the attention operator [4][5]. - The fusion operator optimization combines multiple operators into a single composite operator, enhancing parallel processing and reducing redundant data movement, leading to substantial performance improvements in model inference [7][9]. - SMTurbo technology enables ultra-low latency memory sharing across 384 cards, achieving sub-microsecond delays and enhancing memory access throughput by over 20% in cross-machine communication scenarios [10][9]. Group 2: Future Developments - Future research on AMLA will focus on optimizing the MLA operator for quantization scenarios, expanding its application [12]. - The fusion operator optimization will explore its application across more model architectures, promoting efficient inference of large language models on Huawei's Ascend hardware [12]. - Load/Store optimization will balance read and write loads, aiming for practical benefits in large batch sizes within Deepseek dispatch and combine scenarios [12].