Core Insights - Huawei is positioning itself as a leader in AI computing power, directly competing with Nvidia by unveiling a comprehensive AI computing architecture at the Huawei Connect Conference [1][2] - The company has ambitious plans for its Ascend chip series, with multiple new chips scheduled for release between 2026 and 2028, indicating a long-term commitment to enhancing AI capabilities [1][3] - Huawei's new supernodes and superclusters, including Atlas 950 and Atlas 960, are touted as the world's strongest computing clusters, with capabilities exceeding 500,000 and 1 million cards respectively [2] Group 1: AI Computing Power - Huawei's Ascend chip roadmap includes the release of Ascend 950PR in Q1 2026, Ascend 950DT in Q4 2026, Ascend 960 in Q4 2027, and Ascend 970 in Q4 2028, showcasing a strategic long-term vision for AI chip development [1][3] - The company emphasizes the importance of computing power as a key driver for artificial intelligence, particularly in the context of China's AI development [1] Group 2: Supernodes and Superclusters - The Atlas 950 SuperPoD and Atlas 960 SuperPoD supernodes support 8,192 and 15,488 Ascend cards respectively, setting new benchmarks in computing power, memory capacity, and interconnect bandwidth [1][2] - Huawei has deployed over 300 CloudMatrix 384 supernodes, serving more than 20 clients, indicating a strong market presence and customer adoption [2] Group 3: Interconnect Technology - Huawei has developed a new interconnect protocol called Lingqu, aimed at addressing the challenges of large-scale supernode interconnectivity, and plans to open the Lingqu 2.0 technology specifications to foster an open ecosystem [4]

Core Viewpoint - The A-share market experienced a significant rise on September 18, with the electronic and automotive sectors leading the surge, indicating strong investor interest and potential growth in these industries [1][4][13]. Market Overview - The Shanghai Composite Index approached 3900 points, closing up 0.45%, while the Shenzhen Component and ChiNext Index rose by 0.79% and 0.49%, respectively [1]. - The Sci-Tech 50 Index saw a notable increase of over 4.6% during the day, closing up 3.4% with a trading volume of 829.64 billion CNY [2][3]. Sector Performance - The electronic, communication, automotive, and machinery equipment sectors showed strong performance, with significant activity in concepts like optical modules, semiconductor equipment, and advanced packaging [4][6]. - Over 2700 stocks in the market rose, with total trading volume reaching 1.72 trillion CNY, indicating increased market activity compared to the previous day [3]. Electronic Sector Highlights - The electronic sector saw a collective rise, particularly in chip and computing power stocks, with several companies hitting the daily limit up, including Huicheng Co., Heertai, and others [7][11]. - Industrial Fulian reached a new historical high of 67.23 CNY per share, with a total market capitalization of 1.34 trillion CNY [8][9]. Automotive Sector Highlights - The automotive sector also experienced a surge, with multiple stocks, including Suoao Sensor and Wanxiang Qianchao, hitting the daily limit up [14][15]. - The Ministry of Industry and Information Technology has been actively promoting standards for intelligent connected vehicles, which is expected to enhance industry growth and regulatory compliance [16]. - The passenger car market in China showed robust performance, with sales reaching 18.35 million units from January to August 2025, marking a 14% year-on-year increase [17].

Core Insights - The A-share AI computing stocks experienced a rapid surge, leading the market, with the 5G Communication ETF (515050) rising nearly 3% as of 10:55 AM on September 18 [1] - Huawei's rotating chairman emphasized the importance of computing power for artificial intelligence, sharing future plans for Ascend chips, including the launch of Ascend 950PR in Q1 2026 and Ascend 970 in Q4 2028 [1] - The 5G Communication ETF has attracted over 260 million yuan in investments over the past two days, with a total scale exceeding 8 billion yuan, focusing on the supply chains of Nvidia, Apple, and Huawei [1] Company and Industry Summary - The 5G Communication ETF (515050) is heavily weighted towards optical communication stocks (42%) and PCB circuit board stocks (14%) [1] - Key stocks such as Guangku Technology surged over 17%, Hengtong Optic-Electric hit the daily limit, and Industrial Fulian rose nearly 8%, with the total market capitalization surpassing 1.3 trillion yuan [1]

Core Insights - The report from Guohai Securities highlights the increasing demand for AI computing power driven by large model training, with new architectures like GB300 and Vera Rubin expected to emerge, benefiting various segments of the computing power supply chain [1] Group 1: GPU Developments - The GB300 GPU, based on the Blackwell Ultra architecture, shows a significant performance increase with a floating-point computing capability of 15 PFLOPS, which is 1.5 times that of the B200 [1] - Vera Rubin's NVL144 performance is 3.3 times that of GB300's NVL72, while Rubin Ultra NVL576 is projected to outperform GB300 NVL72 by 14 times, with memory capacity being 8 times greater [1] - NVIDIA's revenue for FY2026 Q2 is expected to reach $46.7 billion, reflecting a year-on-year growth of 56% [1] Group 2: Server Architecture - The GB300 NVL72 system consists of 18 compute trays and 9 switch trays, integrating 72 Blackwell Ultra GPUs and 36 Grace CPUs, with potential performance improvements of up to 50 times compared to Hopper [1] - The GB300 NVL72 integrates ConnectX-8 network cards and BlueField-3 DPUs, while the Rubin Ultra NVL576, set to launch in 2027, will utilize the Kyber architecture for enhanced density [1] Group 3: Networking Innovations - CPO technology replaces traditional pluggable optical modules, enhancing energy efficiency by 3.5 times and deployment speed by 1.3 times [2] - Quantum-X and Spectrum-X switches reduce reliance on traditional optical transceivers, providing up to 400 Tbps throughput for large-scale AI factories [2] - The new NVSwitch 7.0 expands to connect 576 GPUs, enabling non-blocking communication for larger GPU interconnections [2] Group 4: HBM Market Trends - HBM4 is expected to enter mass production in 2026, with SK Hynix leading the market [2] - SK Hynix delivered the world's first 12-layer stacked HBM4 samples to major clients in March 2025, while Samsung plans to provide samples to AMD and NVIDIA by July 2025 [2] - Micron is also showcasing plans for next-generation HBM4 in its AI memory product lineup, with mass production anticipated in 2026 [2] Group 5: Cooling Technologies - The GB300 NVL72 employs a full liquid cooling solution with independent liquid cooling plates for each chip, enhancing heat dissipation efficiency [2] - Liquid cooling technology, including both cold plate and immersion cooling, is noted for its higher efficiency and lower operational costs, with cold plate cooling being more mature and widely adopted [2]

在2025华为全联接大会上，华为轮值董事长徐直军再次强调：算力不仅是人工智能发展的关键，更是中 国人工智能自主可控的关键。 徐直军还首次对外披露了昇腾系列AI芯片的未来规划： 预计 2026年第一季度推出昇腾950PR，第四季度推出昇腾950DT，2027年第四季度推出昇腾960芯片， 2028年第四季度推出昇腾970。值得关注的是，950PR采取了华为自研HBM。 这条时间表意味着，华为将以基本每年迭代一次的节奏，持续推进昇腾的演进， 意在构建一个可持 续、逐步逼近甚至替代国际领先水平的算力底座，这与英伟达、AMD的AI芯片迭代速度保持一致。 华为昇腾的发展历程经历了几个关键节点：2018年10月在华为全联接大会上，徐直军提出AI战略并发 布Ascend系列IP和处理器。 2018年11月发布全球首个AI芯片昇腾310。2019年8月：昇腾910芯片正式推 出，标志着华为AI芯片进入实际应用阶段。 由于美国制裁影响，昇腾经历了一段时间沉寂。昇腾920芯片于2025年4月10日正式发布 特别是2025年6月华为推出昇腾384超节点，算力达300PFLPS，超过英伟达同类产品。 无疑昇腾的迭代路线不仅是一份产品 ...

Group 1 - The electronic sector is leading the market with the electronic ETF (515260) surging over 4.6%, reaching a new high since its launch [1] - Key stocks in the electronic ETF include RichChip, which hit the daily limit, and companies like Zhongwei, which rose over 14%, and Northern Huachuang, which increased by more than 8% [1] - The electronic ETF focuses on semiconductor and consumer electronics, benefiting from positive catalysts in both areas [2] Group 2 - In the semiconductor sector, Huawei's prediction of a tenfold increase in total computing power by 2035 and a 500-fold increase in storage demand is driving growth in the semiconductor industry [2] - The AI-driven demand for computing power is expected to attract more investment into core technology and innovation areas within the semiconductor sector [2] - In the consumer electronics space, the upcoming Meta Connect conference and Apple's recent product launch are seen as significant events that could boost related stocks [3] Group 3 - The electronic ETF's composition shows that 43.34% of its holdings are related to Apple's supply chain, indicating a strong correlation with Apple's performance [3] - The electronic sector has attracted significant capital inflow, with 213 billion yuan in main funds as of the report, and substantial inflows over the past 5, 20, and 60 days [3] - The electronic industry is currently in an innovation phase, with expectations for breakthroughs in terminal innovation and profit releases in the future [4]

人工智能浪潮席卷全球，算力需求激增，机房基础设施面临前所未有的挑战，如何构建新一代AIDC （人工智能数据中心）已成为产业共同面对的核心命题。 9月17日，在首届AIDC产业发展大会上，科研机构、数据中心服务商、互联网运营商代表等产业链各方代表齐 聚上海，围绕标准共建、技术共研与生态协同等话题讨论，并在大会上预发布《AIDC基础设施规范》，为未 来2～3年的AIDC建设提供参考，推动产业高质量发展。 值得一提的是，华为新发布的面向智能世界2035十大技术趋势报告披露，到2035年全社会的算力总量将增长10 万倍，在计算架构、材料器件、工程工艺、计算范式四大核心层面实现颠覆性创新，最终催生新型计算的全面 兴起。 报告指出，随着大模型的发展，AI智能体将从执行工具演进为决策伙伴，驱动产业革命。到2035年，全社会 的算力总量将增长10万倍，计算领域将突破传统冯•诺依曼架构的束缚，在计算架构、材料器件、工程工艺、 计算范式四大核心层面实现颠覆性创新，最终催生新型计算的全面兴起。 全球计算联盟（GCC）秘书处CTO苗福友接受券商中国记者采访时表示，随着大模型进入实用化阶段，人工 智能正催生出大量全新的应用场景，AI产 ...

Core Viewpoint - The article discusses the unprecedented challenges faced by data center infrastructure due to the explosive growth in computing power demand driven by artificial intelligence (AI) technologies, emphasizing the need for a new generation of Artificial Intelligence Data Centers (AIDC) [1][4]. Group 1: AIDC Development and Standards - The first AIDC Industry Development Conference was held in Shanghai, where representatives from various sectors discussed standardization, technology collaboration, and ecosystem synergy, leading to the pre-release of the "AIDC Infrastructure Specification" to guide AIDC construction over the next 2-3 years [2][8]. - The AIDC infrastructure standard aims to address the challenges of planning, delivery, and operation in the rapidly evolving AIDC industry, promoting standardized and scalable development [8][9]. Group 2: Computing Power Demand - The demand for computing power is experiencing explosive growth as AI technology penetrates various industries, becoming a key driver of technological revolution and industrial transformation [4][5]. - The compound annual growth rate (CAGR) for domestic AI servers is over 20%, while the scale of intelligent computing power is growing at over 30%, indicating a need for continuous upgrades in AI infrastructure [5][6]. Group 3: Challenges in AIDC Construction - AIDC construction faces significant challenges related to heat, electricity, and space, as well as a lack of standards and long construction cycles, which complicate the establishment of AI data centers [8][9]. - The industry is currently in a phase of exploration, leading to cost inefficiencies and extended delivery times, with data center construction often taking years, which delays the deployment of AI equipment [8][9]. Group 4: Future Trends and Innovations - By 2035, the total computing power in society is projected to increase by 100,000 times, with disruptive innovations expected in computing architecture, materials, engineering processes, and computing paradigms [5][6]. - The emergence of large models in AI is expected to create numerous new application scenarios, marking a period of rapid application explosion in the AI industry [6][7].

Group 1: AI Model Development - Elon Musk announced on X platform that the Grok 5 model will begin training in a few weeks, expressing optimism that xAI could achieve AGI through Grok 5, a thought he had not considered before [1] - The computational power invested in the Grok series models has significantly increased with each generation, with Grok 4's reinforcement learning (RL) computational power being 10 times greater than that of Grok 3, and RL power now exceeding pre-training computational power [1] - It is anticipated that by 2025, model iterations will accelerate, with leading companies releasing new models every two months, pushing the limits of model performance [1] Group 2: Industry Trends - The rapid iteration of model vendors is expected to have a significant impact on the global AI industry, particularly in areas such as computational infrastructure, driven by a competitive arms race among tech giants in the frontier model field, explosive growth in computational power due to scaling laws, and enhanced reasoning capabilities unlocking more application scenarios [1] Group 3: Company Developments - Sanbian Technology announced plans to establish a wholly-owned subsidiary in the United States in 2024, with their main transformer being integrated into Musk's xAI supercomputer center [2] - Changxin Bochuang's self-developed AOC optical transceiver chips hold a leading global market share, with their 800G multi-mode AOC for data centers having completed multi-platform verification [2]

1. Report Industry Investment Rating - The investment rating for the global economy in the macro and financial sector is (Bullish) [1] 2. Core View of the Report - Overseas funds are continuously flowing into Chinese assets, leading to the rise of popular Chinese concept stocks and new highs in relevant indices. Huawei's report on the future trends of key technologies shows the potential of AGI and AI. The consumption of enterprise - level large - model Tokens in China has increased significantly, and Alibaba has advantages. Ningde Times has high - tech production lines and great potential in the energy storage business. The UK and the US have reached a technology agreement, and the US may set up a mining investment fund. The gold price increase is driven by European and American investors. China's economic policies and international economic trends are also favorable [1] 3. Summary by Related Catalog 3.1 Important Information - Overseas funds are flowing into Chinese assets, with the Nasdaq Golden Dragon China Index reaching a new high since February 2022 and the Hang Seng Tech Index hitting a new high since March [1] - Huawei released the Smart World 2035 series of reports, stating that AGI will be the most transformative driving force in the next decade, and the total computing power of the whole society will increase by 100,000 times in 2035 [1] - The daily average Token consumption of Chinese enterprise - level large models reached 10.2 trillion in the first half of 2025, a surge of 363% compared to the second half of 2024. Goldman Sachs believes Alibaba is in an advantageous position [1] - Morgan Stanley thinks Ningde Times' self - developed production line has high - tech features, with a current utilization rate of over 90%, and it is building new capacity with a target of 1TWh next year [1] - The UK and the US reached a "Technology Prosperity Agreement", and US tech giants will invest over £31 billion in AI and other projects [1] - The rise in the gold price is mainly driven by European and American investors, with significant increases in the US trading session and ETF holdings [1] - The US International Development Finance Corporation is in talks to set up a mining investment fund with a scale of about $5 billion [1] 3.2 Global Economic Logic - China is implementing the "Artificial Intelligence +" action, and the energy storage installation scale will double in three years. In August, a net inflow of $39 billion entered Chinese bonds and stocks [1] - The Federal Reserve has entered an interest - rate cut cycle, and the US Court of Appeals ruled that "reciprocal tariffs" are illegal [1] - US capital goods imports in July reached a record high of $95.8 billion, and the manufacturing industry is accelerating. The eurozone's manufacturing PMI in August exceeded the boom - bust line for the first time since June 2022 [1] - Huawei released the Smart World 2035 series of reports, and Yushu Technology open - sourced a world model - motion architecture for multiple types of robots. Global AI giants are starting a model competition [1]