Core Insights - The article discusses the collaboration and solutions in the supernode field, highlighting the major players and their respective strategies in the market [3][4]. Supernode Collaboration and Solutions - Major CSP manufacturers are seeking customized server cabinet products from server suppliers, with a focus on NV solutions [4]. - Key supernode solutions in China include Tencent's ETH-X, NV's NVL72, Huawei's Ascend CM384, and Alibaba's Panjiu, which are either being promoted or have existing customers [4]. - ByteDance is planning an Ethernet innovation solution for large models, primarily based on Broadcom's Tomahawk, but it has not yet been promoted [4]. - Tencent's ETH-X collaborates with Broadcom and Amphenol, utilizing Tomahawk switches and PCIe switches for GPU traffic management [5]. - The main applications of these solutions differ: CM384 focuses on training and large model computation, while ETH-X is more inclined towards inference [5]. Market Share and Supplier Landscape - The supernode solutions have not yet captured a significant market share, with traditional AI servers dominated by Inspur, H3C, and others [6]. - From September 16, CSPs including BAT were restricted from purchasing NV compliant cards, leading to a shift towards domestic cards, which are expected to reach 30%-40% in the coming years [6]. - The overseas market share for major internet companies like Alibaba and Tencent remains small, with ByteDance's overseas to domestic ratio projected to improve [6]. Vendor Competition and Second-Tier Landscape - Inspur remains competitive in terms of cost and pricing, while the competition for second and third places among suppliers is less clear [8]. - The second-tier internet companies have smaller demands, and mainstream suppliers are not actively participating in this segment [9]. - The article notes that the domestic AI ecosystem is lagging behind international developments, with significant advancements expected by 2027 [9][10]. Procurement and Self-Developed Chips - Tencent and Alibaba have shown a preference for NV cards when available, with a current ratio of NV to domestic cards at 3:7 for Alibaba and 7:3 for ByteDance [10]. - The trend towards supernodes is driven by the need for increased computing power and reduced latency, with expectations for large-scale demand in the future [10]. Economic and Technical Aspects - The article highlights the profit margins for AI servers, with major manufacturers achieving higher gross margins compared to general servers [11]. - The introduction of software solutions is expected to enhance profitability, with significant profit increases anticipated from supernode implementations [11].

Investment Rating - The report indicates a positive investment outlook for the ASIC industry, highlighting significant growth potential driven by increasing demand for AI applications and specialized chip designs [2]. Core Insights - The report emphasizes the distinct business models of ASIC and GPU, noting that ASICs are specialized chips tightly coupled with specific downstream applications, while GPUs are general-purpose chips [3][10]. - ASICs demonstrate superior cost-effectiveness and efficiency, with notable examples such as Google's TPU v5 achieving 1.46 times the energy efficiency of NVIDIA's H200, and Amazon's Trainium2 reducing training costs by 40% compared to GPU solutions [3][15]. - The report forecasts that the global AI ASIC market could reach $125 billion by 2028, with significant contributions from major players like Broadcom and Marvell [30]. Summary by Sections 1. AI Model Inference Driving ASIC Demand - The global AI chip market is projected to reach $500 billion by 2028-2030, with AI infrastructure spending expected to hit $3-4 trillion by 2030 [8]. - ASICs are recognized for their strong specialization, offering cost and efficiency advantages over GPUs, particularly in AI applications [9][14]. 2. High Complexity of ASIC Design and Value of Service Providers - ASIC design involves complex processes requiring specialized service providers, with Broadcom and Marvell being the leading companies in this space [41][42]. - The report highlights the importance of design service providers in optimizing performance and reducing time-to-market for ASIC products [55][60]. 3. Domestic Developments: Not Just Following Trends - Domestic cloud giants like Alibaba and Baidu have made significant strides in ASIC self-research, establishing independent ecosystems rather than merely following international trends [4][30]. - The report identifies key domestic design service providers such as Chipone, Aojie Technology, and Zhaoxin, which are well-positioned to benefit from the growing demand for ASICs [41]. 4. Key Trends in Domestic ASIC Development - The report identifies PD separation and supernode architectures as two core trends in domestic ASIC development, with companies like Huawei and Haiguang leading the way [4][30]. - These trends reflect a shift towards more flexible and efficient chip designs that cater to diverse industry needs [4]. 5. Valuation of Key Companies - The report includes a valuation table for key companies in the ASIC sector, indicating strong growth prospects and market positioning for firms like Broadcom and Marvell [5].

Investment Rating - The report maintains a positive outlook on the ASIC industry, indicating a favorable investment rating for the sector [2]. Core Insights - The report highlights the significant cost-effectiveness and efficiency advantages of ASICs over GPUs, particularly in the context of AI model inference, with Google's TPU v5 demonstrating an energy efficiency ratio 1.46 times that of NVIDIA's H200 [3][19]. - The increasing penetration of AI applications is driving a surge in inference demand, expanding the market for ASICs, with projections indicating the global AI ASIC market could reach $125 billion by 2028 [3][32]. - The report emphasizes the complexity of ASIC design, underscoring the critical role of design service providers like Broadcom and Marvell, which are expected to benefit from the growing demand for custom ASIC solutions [4][44]. Summary by Sections 1. Demand Driven by Large Model Inference - The global AI chip market is projected to reach $500 billion by 2028-2030, with significant growth in AI infrastructure spending anticipated [13]. - ASICs are specialized chips that offer strong cost and efficiency advantages, particularly in specific applications like text and video inference [14][19]. - The report notes that the demand for ASICs is expected to rise sharply due to the increasing consumption of tokens in AI applications, exemplified by the rapid growth of ChatGPT's user engagement [25][31]. 2. High Complexity of ASIC Design and Value of Service Providers - ASIC design involves a complex supply chain, with cloud vendors often relying on specialized design service providers for chip architecture and optimization [41][44]. - Broadcom's ASIC revenue is projected to exceed $12 billion in 2024, driven by the success of its TPU designs for Google and other clients [60]. - The report identifies the importance of a complete IP system and design experience as key factors for service providers to secure new orders in the ASIC market [63]. 3. Domestic Developments: Not Just Following Trends - Leading Chinese cloud providers like Alibaba and Baidu are making significant strides in self-developed ASICs, indicating a robust domestic ecosystem [3][4]. - The report highlights the emergence of domestic design service providers such as Chipone and Aowei Technology, which are positioned to capitalize on the growing demand for ASICs [3][4]. - The trends of PD separation and supernodes are identified as critical developments in the domestic ASIC landscape, with companies like Huawei and Haiguang leading the way [4][44]. 4. Key Trends in Domestic ASIC Development - PD separation involves using different chips for prefill and decode tasks, enhancing efficiency in specific applications [4]. - Supernodes are being developed to create unified computing systems through high-bandwidth interconnections, with early implementations seen in domestic companies [4][44].

Core Viewpoint - The article discusses the advancements in computing power architecture, particularly focusing on Alibaba Cloud's new supernode design and its implications for large model training and inference in the AI sector [4][10]. Group 1: Supernode Design and Technology - Alibaba Cloud's supernode architecture addresses the increasing demands for memory capacity and bandwidth in large model training, moving beyond traditional GPU setups [4]. - The supernode design leverages the advantages of PPU chip design, emphasizing high-density integration [6]. - The supernode can support up to 64 cards in a single machine, with a power requirement of 300 kilowatts, necessitating advanced interconnect protocols [9]. Group 2: UALink Protocol and Industry Collaboration - The UALink protocol, initiated by a consortium including AMD, AWS, and others, aims to enhance interconnectivity in computing systems, with Alibaba Cloud as a member [5]. - The UALink alliance was formed to address the high costs of evolving proprietary technologies in the industry, with AMD contributing its Infinity Fabric protocol [5]. Group 3: PPU Specifications and Performance - The PPU features 96GB of HBM2e memory, surpassing the A800's 80GB and matching the H20's capacity, with an inter-chip bandwidth of 700GB/s [10]. - The PPU supports PCIe 5.0×15 interfaces, which is an improvement over the A800's PCIe 4.0×16, while maintaining a power consumption of 400W [10]. - The PPU is available in two versions, with the base version achieving a peak performance of 120 TFLOPS, focusing on AI inference tasks [10].

Core Insights - Intel and Nvidia have formed a historic partnership, with Nvidia investing $5 billion in Intel to co-develop customized data center and personal computing products, aiming to enhance large-scale computing capabilities [1][2] - This collaboration signifies a shift in the semiconductor industry, potentially leading to market differentiation, where competitors like AMD and ARM may face increased pressure [1][11] Group 1: Partnership Details - Nvidia will utilize its NVLink technology to integrate its AI and accelerated computing strengths with Intel's advanced CPU technology, providing cutting-edge solutions for clients [2][4] - Intel will customize x86 processors for Nvidia's AI infrastructure and launch a new x86 system-on-chip (SoC) that integrates Nvidia's RTX GPU for various PC products [2][4] Group 2: Market Reactions - Following the announcement, Intel's stock surged nearly 30%, closing with a 22.77% increase at $30.57 per share, while Nvidia's stock rose 3.49% to $176.24 per share [6][9] - The partnership has raised concerns for competitors AMD and ARM, with AMD's stock dropping over 5% initially, reflecting market apprehension about the new alliance [9][10] Group 3: Strategic Implications - Nvidia's investment is seen as a strategic move to solidify its position in the CPU market while mitigating risks from competitors like Microsoft and Amazon, which are developing their own chips [4][10] - The collaboration may also challenge TSMC if Nvidia shifts some of its chip manufacturing to Intel, although TSMC's market outlook remains stable for now [8][10] Group 4: Impact on Chinese Semiconductor Industry - The partnership could further entrench the U.S. dominance in high-end computing and data center chips, complicating competition for Chinese firms [11][12] - Chinese semiconductor companies are expected to accelerate their independent innovation efforts, particularly in the development of "super nodes" to enhance their competitive edge [11][13]

Core Viewpoint - Intel and Nvidia have formed a historic partnership, with Nvidia investing $5 billion in Intel to co-develop customized data center and personal computing products, marking a significant shift from competition to collaboration in the semiconductor industry [2][3][4]. Group 1: Partnership Details - Nvidia will utilize its NVLink technology to seamlessly integrate its AI and accelerated computing advantages with Intel's advanced CPU technology, aiming to provide cutting-edge solutions for customers [3]. - Intel will customize x86 processors for Nvidia's AI infrastructure and launch x86 system-on-chip (SoC) products that integrate Nvidia's RTX GPU chiplets for various PC applications [4]. - This partnership is seen as a strategic move for both companies, with Nvidia seeking to enhance its CPU market presence and Intel aiming to revitalize its position in the semiconductor industry [5][6]. Group 2: Market Impact - Following the announcement, Intel's stock surged nearly 30%, closing with a 22.77% increase, while Nvidia's stock rose by 3.49% [6]. - The collaboration is expected to create a new competitive landscape in the semiconductor market, particularly affecting AMD and ARM, which may face increased pressure due to the strengthened position of Intel and Nvidia [9][10]. - The partnership may also challenge TSMC's dominance if Nvidia shifts some of its chip manufacturing to Intel [9]. Group 3: Implications for Chinese Semiconductor Industry - The alliance between Nvidia and Intel could further consolidate the U.S. dominance in high-end computing and data center chips, complicating competition for Chinese manufacturers [11]. - Chinese companies are likely to accelerate their independent innovation efforts, focusing on developing "super nodes" as a key strategy to enhance their competitive edge [11][12]. - Domestic firms are exploring distributed architectures and interconnect innovations to create independent systems that do not rely on Nvidia's ecosystem, which is crucial for China's long-term technological independence [13].

Investment Rating - The report suggests a positive outlook for the industry, indicating that the performance of related stocks is expected to outperform the benchmark index over the next 12 months [17]. Core Insights - Huawei's roadmap for AI chips, supernodes, and computing clusters is anticipated to lead a new paradigm in China's AI infrastructure, with supernodes becoming the new norm in AI infrastructure construction [5][11]. - The domestic computing power is constrained by manufacturing processes, but the development of "supernode + cluster" computing solutions is expected to continuously meet computing power demands [5][11]. - The introduction of supernodes is expected to enhance demand and value across multiple computing segments, including increased interconnectivity, liquid cooling value, and the transition from traditional product manufacturers to system solution providers [5][11]. Summary by Sections Event Description - On September 18, 2025, during the Huawei Connect 2025 conference, Huawei announced its AI chip roadmap, including the launch of the Ascend 950PR chip in Q1 2026, the Ascend 950DT chip in Q4 2026, the Ascend 960 chip in Q4 2027, and the Ascend 970 chip in Q4 2028 [8]. Event Commentary - Huawei's recent announcements include the launch of the Atlas 950 SuperPoD and Atlas 960 SuperPoD, supporting 8192 and 15488 Ascend cards respectively, and the introduction of new supernode clusters, Atlas 950 SuperCluster and Atlas 960 SuperCluster, with computing power exceeding 500,000 and reaching one million cards [11]. - The report emphasizes that supernodes are rapidly becoming a new standard in AI infrastructure, with Huawei leveraging its communication capabilities to overcome key bottlenecks and support large model training and inference [11]. - The domestic semiconductor industry is accelerating the iteration of domestic technologies, with improvements in advanced manufacturing processes and increasing localization of supporting equipment and materials [11]. Investment Opportunities - The report suggests focusing on investment opportunities in the following areas: leading domestic AI chip companies like Cambricon, high-end CPU and DCU leaders, supernode server manufacturers such as FiberHome and Digital China, supernode-related partners of Huawei, and suppliers in the advanced semiconductor manufacturing chain [11].

Investment Rating - The report maintains an "Outperform" rating for the industry [6] Core Insights - Huawei's new products, including Ascend chips and supernodes, are set to lead a new era in computing power, with a clear roadmap for product iterations from 2025 to 2028 [1][14] - The introduction of self-developed HBM technology marks a significant advancement in memory bandwidth, enhancing the efficiency of large model training [3][23] - The supernode architecture is designed to integrate hundreds of processors, reshaping the competitive landscape in AI infrastructure [22][24] Summary by Sections 1. Huawei Computing Power Product Launch - Ascend chips will follow a yearly iteration schedule, with the 910C launching in Q1 2025 and subsequent models (950PR, 950DT, 960, 970) planned through 2028 [14][15] - The Ascend 950 series introduces low-precision data formats and self-developed HBM, enhancing training efficiency for large models [15][16] - The Ascend 960 and 970 are expected to double performance metrics across various parameters, including computing power and memory bandwidth [18][19] 2. Supernode Products - The supernode data center (Atlas 900/950/960 SuperPoD) is designed for large-scale AI training, achieving EFLOPS-level computing power with high bandwidth and low latency [2][27] - The supernode cluster (Atlas 950/960 SuperCluster) enhances network performance and energy efficiency, reaching ZFLOPS-level computing power [2][37] - The enterprise-grade air-cooled supernode server (Atlas 850) is tailored for post-training and multi-scenario inference, supporting flexible scaling [2][38] 3. Related Investment Targets - Key investment targets include hardware partners for Ascend, domestic wafer foundries, and companies involved in copper connections, optical connections, power supplies, PCBs, and cooling solutions [4][46]

Core Insights - Huawei is making significant advancements in AI computing power, showcasing a comprehensive AI computing architecture at the Huawei Connect Conference [1][3] - The company aims to establish a sustainable computing ecosystem based on its proprietary Ascend chips and Kunpeng CPUs, positioning itself as a key player in the AI market [5][8] Group 1: AI Computing Strategy - Huawei's AI computing strategy is centered around the Ascend chip series, with a roadmap for iterative upgrades over the next three years, aiming for a doubling of computing power annually [3][5] - The upcoming Ascend 950 series is designed to enhance training efficiency and support various low-precision data formats, indicating a focus on optimizing AI workloads [3][4] Group 2: Supernodes and Clusters - Huawei introduced new supernode products, Atlas 950 SuperPoD and Atlas 960 SuperPoD, which support 8192 and 15488 Ascend cards respectively, positioning them as leaders in global computing power [6][7] - The Atlas 950 SuperCluster and Atlas 960 SuperCluster are set to exceed 500,000 and 1 million cards, respectively, further solidifying Huawei's dominance in AI infrastructure [6] Group 3: Competitive Landscape - Despite NVIDIA's current dominance in the AI chip market, Huawei is rapidly developing its Ascend and Kunpeng chips to compete effectively within the Chinese AI landscape [5][6] - Huawei's approach includes a focus on system-level engineering and innovation, aiming to create a competitive edge in the evolving AI computing sector [7][9]

Core Viewpoint - Huawei is leading a new paradigm in AI infrastructure with its innovative supernode interconnection technology, emphasizing the importance of computing power in artificial intelligence development [2][8]. Summary by Sections Ascend Chip Development - Huawei has committed to the monetization of Ascend hardware and plans to open-source its CANN compiler and virtual instruction set interface by December 31, 2025 [2]. - The Ascend 950 series, including Ascend 950PR and Ascend 950DT, is set to significantly enhance computing power and efficiency compared to previous models [3][4]. - The Ascend 950 chip will support new data formats and achieve a computing power of 1 PFLOPS for FP8 and 2 PFLOPS for FP4, with interconnect bandwidth increased to 2 TB/s [5][6]. Upcoming Chip Releases - The Ascend 950PR chip is targeted for Q1 2026, focusing on inference prefill and recommendation scenarios, while the Ascend 950DT will be released in Q4 2026, emphasizing inference decode and training [6][7]. - The Ascend 960 chip, expected in Q4 2027, will double the specifications of the Ascend 950, enhancing performance for training and inference [7]. - The Ascend 970, planned for Q4 2028, aims to further upgrade performance metrics across various specifications [7]. Supernode Products - The Atlas 950 supernode, based on the Ascend 950DT, will support 8192 Ascend 950DT chips, achieving FP8 computing power of 8 EFLOPS and interconnect bandwidth of 16 PB/s, set to launch in Q4 2026 [9][10]. - The Atlas 960 supernode, launching in Q4 2027, will support 15488 cards, with FP8 computing power reaching 30 EFLOPS and interconnect bandwidth of 34 PB/s [11]. Interconnection Technology - Huawei has developed a new interconnection protocol, "Lingqu," to support large-scale supernodes, enhancing reliability and bandwidth while reducing latency [18][19]. - The Atlas 950 SuperCluster, consisting of 64 Atlas 950 supernodes, will achieve a total FP8 computing power of 524 EFLOPS, launching alongside the Atlas 950 supernode in Q4 2026 [20]. Future Directions - Huawei aims to continue evolving its supernode and cluster products to meet the growing demands for AI computing power, with a focus on both AI and general computing applications [12][21].