Core Viewpoint - The article discusses Huawei's innovative "super node" architecture, which aims to redefine large-scale effective computing power in AI by addressing the limitations of traditional server architectures and enhancing interconnectivity through the self-developed UnifiedBus protocol [3][4][12]. Group 1: Super Node Architecture - The super node architecture represents a deep restructuring of computing system architecture, moving from a "stacked" model to a "fused" model that allows multiple machines to function as a single device [4][9]. - This architecture aims to eliminate the communication bottlenecks inherent in traditional server setups, where data exchange between servers can lead to significant delays and inefficiencies [5][11]. - Huawei's super node can reduce communication latency to the nanosecond level, significantly improving cluster utilization and lowering communication costs, with the goal of achieving linear scalability of effective computing power [11][12]. Group 2: Product Offerings - Huawei introduced the Atlas 950 SuperPoD and Atlas 960 SuperPoD, which support 8192 and 15488 Ascend cards respectively, showcasing superior performance in key metrics such as card scale, total computing power, memory capacity, and interconnect bandwidth [17][20]. - The Atlas 850, an enterprise-grade air-cooled AI super node server, lowers the barrier for enterprises to adopt super node architecture without requiring complex liquid cooling modifications [21]. - The TaiShan 950 SuperPoD extends the super node architecture to general computing, offering ultra-low latency and memory pooling capabilities beneficial for databases and big data applications [25]. Group 3: Ecosystem Strategy - Huawei emphasizes an ecosystem strategy of "hardware openness and software open-source," encouraging industry partners to engage in secondary development and enrich product offerings based on the UnifiedBus protocol [26][28]. - The company aims to build a unified, scalable computing foundation that provides a consistent, high-performance computing experience across various environments, from cloud to enterprise [28].

Group 1 - Several semiconductor companies announced acquisitions and investments in the chip sector over the weekend [1] - Huahong announced plans to acquire 97.5% of Huali Micro's equity through a combination of issuing shares and cash, along with raising supporting funds [1] - Dongxin Co. plans to invest approximately 211 million yuan in Shanghai Lishuan, increasing its stake to 35.87% [1] Group 2 - SMIC announced plans to issue A-shares to purchase minority stakes in its subsidiary, SMIC North [2] - AI chips are identified as a core area in the AI industry chain, directly influencing the computing power and efficiency of AI systems [2] - The domestic chip self-sufficiency rate in China is steadily increasing, but there is still a need for improvement, especially with the upcoming restrictions on AI chips in the U.S. by 2025 [2] Group 3 - DeepSeek V3.1 was officially released, optimized for the next generation of domestic chips [2] - Shanxi Securities expresses optimism about the next generation of DeepSeek models, highlighting trends in chip performance improvements and enhanced hardware-software collaboration [2] - The performance of the Ascend 910 next-generation chip is expected to approach mainstream international levels, opening significant market space for second-tier GPU manufacturers [2] Group 4 - Companies such as North Huachuang, Zhongwei Company, Chip Source Micro, Tuojing Technology, Shengmei Shanghai, Huahai Qingke, Zhongke Feice, and Jingyi Equipment are recognized as leaders in semiconductor equipment [3]

Core Insights - The rapid development of large models driven by AI demands significant computational power, leading to the emergence of the "SuperPod" as a key solution for efficient AI training [1][2] - The transition from traditional computing architectures to SuperPod technology signifies a shift in the AI infrastructure competition from isolated breakthroughs to a system-level ecosystem [1][5] Industry Trends - The SuperPod, proposed by NVIDIA, represents a Scale Up solution that integrates GPU resources to create a low-latency, high-bandwidth computing entity, enhancing performance and energy efficiency [2][4] - The traditional air-cooled AI servers are reaching their power density limits, prompting the adoption of advanced cooling technologies like liquid cooling in SuperPod designs [2][5] Market Outlook - The market for SuperPods is viewed positively, with many domestic and international server manufacturers selecting it as the next-generation solution, primarily utilizing copper connections [2][4] - Major Chinese tech companies, including Huawei and Xizhi Technology, are actively developing SuperPod solutions, showcasing significant advancements in AI computing capabilities [5][6] Technological Developments - The ETH-X open standard project, led by the Open Data Center Committee, aims to establish a framework for SuperPod architecture, combining Scale Up and Scale Out networking strategies [4] - Companies like Moer Thread are building comprehensive AI computing product lines, emphasizing the need for efficient collaboration among large-scale clusters to enhance AI training infrastructure [6]

Investment Rating - The report initiates coverage on Huafeng Technology with an "Accumulate" rating and a target price of 59.86 RMB per share, based on a 75x PE valuation for 2026 [6][5]. Core Views - Huafeng Technology is positioned as a leader in high-speed connectors in China, benefiting from the increasing demand for short-distance interconnects driven by AI and domestic computing power expansion. The company is gradually releasing production capacity for high-speed line modules developed for major clients, which is expected to lead to sustained performance growth [1][15]. - The report highlights the growth potential in the communications sector, driven by the demand for high-speed interconnects in AI clusters, with a projected market size of 24.1 billion RMB by 2029, growing at a CAGR of 45% from 2025 to 2029 [2][16]. - In the defense sector, the company is expected to benefit from the "14th Five-Year Plan" military budget increase, with a projected 7.2% year-on-year growth in military spending in 2025, enhancing the outlook for defense orders [3][17]. - The industrial segment is anticipated to see stable growth due to the rising penetration of new energy vehicles and the trend towards 800V high-voltage systems, with the high-voltage connector market projected to reach 33.7 billion RMB by 2026, growing at a CAGR of 42% from 2022 to 2026 [3][18]. Summary by Sections Company Overview - Established in 1958, Huafeng Technology is a leading supplier of optical connectors and interconnection solutions in China, focusing on high-speed connectors and system interconnection solutions across communications, defense, and industrial sectors. The company has achieved significant milestones in developing high-speed backplane connectors, breaking the monopoly of foreign leaders in the domestic market [15][25]. Communications Sector - The company is deeply collaborating with major clients to meet the growing demand for high-speed interconnects in AI clusters. The increasing GPU computing power and bandwidth requirements are driving the need for higher signal transmission rates. The domestic high-speed backplane connector market is projected to reach 24.1 billion RMB by 2029, with a CAGR of 45% from 2025 to 2029 [2][16]. Defense Sector - The defense segment focuses on defense connectors and related system interconnection products. With the military budget expected to reach 1.78 trillion RMB in 2025, a 7.2% increase year-on-year, the company is well-positioned to capture growth in defense orders [3][17]. Industrial Sector - The industrial connectors primarily serve the new energy vehicle and rail transportation sectors. The market for high-voltage connectors in new energy vehicles is projected to reach 33.7 billion RMB by 2026, with a CAGR of 42% from 2022 to 2026. The company is also expanding its applications in drone and eVTOL systems [3][18].

Core Viewpoint - The development of models has shifted from early algorithm optimization to deep innovation at the system engineering level, transitioning from a digital era of bit traffic to a Token economy, with daily Token consumption in China rising from hundreds of billions to tens of trillions [1] Group 1: Model Optimization - Huawei has made significant optimizations for DeepSeek, focusing on three main areas to enhance compatibility and support for enterprise applications [3] - The pre-training aspect includes the implementation of DualPipe technology, which has been improved to minimize static memory usage through the introduction of the DualPipe-V solution [6] - At the operator level, Huawei has enhanced execution efficiency with the MRN PO fusion operator and optimized low-latency communication [7] Group 2: System Architecture - Huawei has developed a new architecture for inference called the "super node" architecture, which interconnects multiple GPUs to reduce communication latency and improve training throughput [14] - The Atlas 900 A3 SuperCluster has been designed to enhance cluster computing efficiency and reliability, achieving a training efficiency increase of 2.7 times [15] - The OmniPlacement algorithm has been introduced to optimize resource utilization by dynamically adapting to expert activation data, improving throughput by 10% [19] Group 3: Load Balancing and Efficiency - Huawei has implemented a large-scale expert parallel (large EP) strategy to enhance inference efficiency, achieving a nearly 20-fold increase in the past two months [17] - The company has developed dynamic priority adjustment and communication optimization strategies to address load balancing challenges in expert parallelism [20]

Core Insights - The rapid growth in demand for AI computing power has outpaced the capabilities of traditional computing architectures, necessitating the development of new solutions like the super node architecture [1] - Huawei Cloud's CloudMatrix 384 super node addresses key technical challenges in AI computing, including communication efficiency, memory limitations, and reliability, achieving a computing power scale of up to 300 Pflops, surpassing NVIDIA's NVL72 by 67% [1] - The introduction of distributed inference platforms and innovative technologies such as Elastic Memory Storage (EMS) significantly enhances resource utilization and performance, reducing latency and improving fault detection rates [2] Group 1 - The demand for AI computing power has increased by 10,000 times, while hardware capabilities have only improved by 40 times in the last eight years [1] - The CloudMatrix 384 super node connects 384 cards into a single super cloud server using a new high-speed interconnect bus [1] - The super node features six technical advantages, including MoE affinity and high reliability [1] Group 2 - The distributed inference platform allows for efficient distributed inference with one card acting as one expert, significantly improving MoE computation and communication efficiency [2] - The MatrixLink service consists of two network layers, enabling high-speed interconnection within the super node and low latency communication [2] - The EMS technology decouples memory from computing power, enhancing resource utilization and reducing the first token latency by up to 80% [2]