当地时间12月3日，黄仁勋在华盛顿战略与国际研究中心发出警告：如果美国公司放任华为等中国竞争 对手抢占市场，中国将很快寻求向全球出口其人工智能技术，打造AI版的"一带一路"基础设施倡议。 与此同时，黄仁勋刚结束与特朗普的会面。面对记者询问，他坦言："我们不知道中国还要不要 H200。"这句简短的回应，道出了这位芯片巨头面对政策变局和市场转变的双重困境。 困局 12月的华盛顿寒意渐浓，黄仁勋的行程安排得密不透风。他先与特朗普面对面讨论芯片出口问题，随后 赶往国会山参加参议院银行委员会的闭门会议。 黄仁勋的核心诉求很明确：全力游说不要再搞新一轮的AI芯片出口限制。 他面对的是一项可能将《GAIN AI法案》纳入年度国防授权法案的提案。该法案要求美国芯片公司必须 优先满足本国市场需求，海外订单尤其是对中国的销售将面临更严格限制。 黄仁勋在国会直言不讳："这个法案对美国自己伤害更大。" 他的努力最终取得了部分成功——GAIN AI法案被暂时搁置。表面上看，这场游说他赢了，但背后冷峻 的现实是：中国市场的信任已经难以挽回。 算盘 黄仁勋此次华盛顿之行的焦虑，直接源于对中国市场的深刻担忧。据他估计，中国目前的AI芯片市场 ...

Core Insights - The token economy in both China and the U.S. is heavily reliant on electricity, with each country facing unique challenges in this regard [1][3] - The U.S. is experiencing a power shortage due to outdated generation and grid infrastructure, limiting token production [1][2] - In contrast, China faces high token production costs due to relatively low-efficiency hardware, impacting the overall cost of token generation [1][3] Group 1: U.S. Challenges - Microsoft CEO Satya Nadella emphasized that the real issue is not a shortage of GPUs but a lack of electricity, which restricts token production and monetization [1] - Major U.S. tech companies are in a race for AI infrastructure investment, which has turned into a competition for electricity supply [1][2] - The construction of large-scale data centers in the U.S. is progressing from 1GW to 10GW, with companies like Crusoe targeting significant capacity increases [1][2] Group 2: Infrastructure and Policy - Silicon Valley giants are urging the White House for support in developing infrastructure, particularly the power grid, to match the pace of AI innovation [3] - OpenAI has suggested that the U.S. needs to add 100GW of electricity capacity annually to compete effectively in AI against China [3] - The U.S. added 51GW of power capacity last year, while China added 429GW, highlighting a significant "power gap" [3] Group 3: China's Challenges - China's AI infrastructure is built on domestic chips, which currently have lower efficiency, leading to increased demand for computational power [3][4] - ByteDance's daily token calls have surged from 16.4 trillion in May to 30 trillion in September, indicating a rapid increase in computational needs [3] - The cost of electricity for a major cloud provider in China is estimated at 8-9 billion yuan for 1GW annually, reflecting the high operational costs associated with domestic chip usage [5] Group 4: Efficiency and Cost - The competition in the token economy involves not just hardware but also the software, tools, and the electricity and cooling systems required to operate them [4] - Huawei's CloudMatrix 384 has shown a significant increase in total computational power but at a much higher energy cost compared to NVIDIA's latest offerings [5][6] - The average industrial electricity cost in the U.S. is approximately 9.1 cents per kWh, while certain regions in China have reduced costs to below 4 cents per kWh, indicating a competitive advantage for Chinese data centers [6]

Investment Rating - The report maintains a "Buy" rating for the company, expecting a price increase of over 15% in the next 6-12 months [5][13]. Core Insights - The company achieved a revenue of 4.962 billion RMB in the first three quarters of 2025, representing a year-on-year growth of 4.62%. The net profit attributable to shareholders was 270 million RMB, up 16.33% year-on-year. In Q3 2025 alone, revenue reached 1.917 billion RMB, marking a 22.0% increase year-on-year and a 14.1% increase quarter-on-quarter. The net profit for Q3 was 108 million RMB, showing an impressive year-on-year growth of 85.28% and a quarter-on-quarter increase of 11.44% [3][4]. Revenue and Profit Analysis - The growth in revenue and profit is primarily driven by the demand for high-speed communication connectors, supported by core clients' investments in high-speed network infrastructure and AI computing needs. The gross margin for Q3 2025 was 19.5%, slightly down by 0.45 percentage points year-on-year and 0.08 percentage points quarter-on-quarter, mainly due to the impact of the solar bracket business. The company has improved its cost control, leading to a decrease in sales, management, and R&D expense ratios, which has positively impacted net profit growth [4]. Market Position and Client Relationships - The company has established strong relationships with top-tier clients, including Huawei and Alibaba, as they accelerate the penetration of ultra-node products. The ultra-node architecture enhances computing cluster performance, which increases the demand for connectors. The company has over 30 years of experience in the connector industry, securing long-term partnerships with high-quality clients, which solidifies its market position [4]. Financial Forecast - The company is projected to achieve revenues of 7.118 billion RMB, 8.720 billion RMB, and 10.535 billion RMB for the years 2025, 2026, and 2027, respectively. The net profit attributable to shareholders is expected to be 402 million RMB, 445 million RMB, and 534 million RMB for the same years, with corresponding P/E ratios of 23, 21, and 17 [5][10].

Core Insights - Chinese AI startups are rapidly capturing global market share with innovative products like smart recording devices and AI toys, while also expanding into overseas markets with technologies like Robotaxi [1][4]. Industry Overview - The efficiency of China's manufacturing supply chain, rapid AI iteration, and keen product awareness are driving competitive advantages for domestic companies [2][6]. - The AI hardware sector is experiencing a wave of innovation, with companies like Plaud.AI achieving significant sales milestones and funding rounds for AI toy startups [3][5]. Market Dynamics - The emergence of AI hardware is not just a product innovation race but a transformation of the industrial paradigm, shifting from "cost-effectiveness" to "intelligence" [5][7]. - Chinese companies are leveraging their technological capabilities to redefine standards in new hardware, moving beyond simple manufacturing to creating advanced solutions [4][6]. Technological Advancements - The integration of AI and manufacturing is leading to a new generation of AI hardware products, supported by a robust domestic supply chain that enhances product development speed and market responsiveness [6][7]. - The domestic AI chip industry is witnessing significant growth, with several companies preparing for IPOs and reporting substantial business orders, indicating a shift from experimental products to commercially viable solutions [8][11]. Ecosystem Collaboration - The collaboration among various stakeholders in the AI ecosystem is crucial for overcoming challenges in computing power and establishing a unified architecture for AI applications [14][16]. - Companies are focusing on building open systems and protocols to enhance interoperability and efficiency in AI computing, which is essential for scaling operations [15][16]. Future Outlook - The convergence of capital investment and technological breakthroughs is expected to solidify China's position in the global AI landscape, transitioning from a follower to a leader in AI innovation [17][18].

Core Insights - Huawei's "Super Node + Cluster" strategy represents a critical response to external pressures and aims to establish a robust AI computing infrastructure in China [2][12][25] - The company emphasizes that AI competition fundamentally revolves around computing power, which is essential for advancing artificial intelligence capabilities [3][4] Group 1: Technological Innovations - The "Super Node" concept integrates multiple computing cards and machines into a single logical unit, enhancing overall efficiency and resource utilization [4][6] - Huawei's Atlas 950 and Atlas 960 SuperNodes support 8192 and 15488 Ascend cards respectively, significantly outperforming Nvidia's upcoming NVL144 in terms of scale and computing power [9][11] - The Lingqu interconnect protocol facilitates the collaboration of numerous computing resources, enabling a more efficient and scalable AI computing environment [13][15] Group 2: Strategic Positioning - Huawei's approach to AI computing is driven by the need to overcome limitations in semiconductor manufacturing due to sanctions, focusing on system architecture innovations [6][7] - The company aims to create a self-sufficient AI ecosystem that does not rely on Western technologies, addressing concerns over supply chain vulnerabilities [22][25] - Huawei's commitment to open-source initiatives is intended to foster collaboration within the industry, allowing for broader access to AI capabilities [19][20] Group 3: Market Implications - The deployment of AI computing resources through cloud platforms allows small and medium enterprises to access advanced AI models, democratizing AI technology across various sectors [23] - Huawei's long-term vision includes building a comprehensive AI ecosystem that supports sustainable growth and innovation within the Chinese AI industry [17][25] - The anticipated release of the Atlas 960 SuperNode and the Ascend 970 chip by 2028 is expected to enhance China's competitive position in the global AI landscape [25][26]

Investment Rating - The report maintains an "Outperform" rating for the communication industry, indicating an expected performance exceeding the benchmark index by more than 10% [36]. Core Insights - Huawei's new Ascend roadmap was unveiled at the 2025 Connectivity Conference, showcasing significant advancements in supernode capabilities, with the upcoming 950PR architecture expected to achieve 1P FP8/2P FP4 computing power and 2TB/s interconnect bandwidth [2][11]. - The collaboration between NVIDIA and Intel aims to create a new order in data centers, with NVIDIA acquiring a 4% stake in Intel and both companies working on customized data center and PC products [4][14]. - The report highlights the potential acceleration of domestic AI chip shipments in 2026, driven by Huawei's leadership in the domestic computing sector [11][12]. Summary by Sections Industry Trends - Huawei's Ascend AI chip roadmap indicates a significant leap in domestic computing capabilities, with the 950 series expected to surpass NVIDIA's previous flagship in interconnect bandwidth [2][11]. - The Atlas 950 SuperPoD supernode utilizes a hybrid copper-optical architecture, enhancing cost-effectiveness and potentially setting a benchmark for domestic computing cluster design [3][12]. Market Overview - The overall market saw an increase during the week of September 22-26, 2025, with the STAR Market index rising by 6.47% and the ChiNext index by 1.96% [6][15]. - The top-performing sectors included liquid cooling (+7.16%), IoT (+5.95%), and IDC (+1.54%) [15]. Stock Performance - Leading stocks included Cambridge Technology (+18.77%), Inspur Information (+11.86%), and Zhongtian Technology (+9.08%) [15][29]. - Stocks with the largest declines were Bochuang Technology (-14.77%), Changfei Fiber (-14.61%), and Yihua Technology (-8.95%) [15][29]. Companies to Watch - Key companies to monitor include Cambricon, Haiguang Information, and ZTE Corporation in the domestic computing sector, as well as Inspur Information and Ziguang Corporation in the supernode market [15].

Investment Rating - The report recommends attention to domestic AI computing targets, indicating a positive outlook for the industry [6][7]. Core Insights - The report highlights Huawei's AI chip roadmap, with the Ascend 950PR chip set to launch in Q1 2026 and the Ascend 970 in Q4 2028. The Ascend 910C was launched in Q1 2025, followed by the Ascend 950PR and 950DT in 2026, and the Ascend 960 in Q4 2027 [6][7]. - Huawei's Ascend 970 chip will feature a computing power of 4 PFLOPS (FP8) and 8 PFLOPS (FP4), with an interconnect bandwidth of 4 TB/s and HBM memory capacity of 288 GB, which is competitive compared to NVIDIA's Blackwell Ultra GB300 [6][7]. - The Atlas 950 SuperPoD, expected to launch in Q4 2025, will be the world's strongest supernode, with over 300 units deployed by September 2025, serving more than 20 clients [7][8]. - Huawei's strategy includes hardware openness and software open-source initiatives, allowing industry self-development and integration into existing systems [8]. Summary by Sections - **AI Chip Roadmap**: Huawei's roadmap includes the launch of the Ascend 950PR in Q1 2026 and the Ascend 970 in Q4 2028, with significant advancements in chip technology and performance metrics [6][7]. - **Supernode Development**: The Atlas 950 SuperPoD and Atlas 960 SuperPoD will support a large number of Ascend cards, leading in various performance metrics and expected to dominate the market for years [7][8]. - **Open Technology Strategy**: Huawei is committed to an open technology approach, providing access to supernode technology and components for industry partners, fostering innovation and development [8].

Core Viewpoint - Computing power is the engine of the digital economy, becoming a core resource in the intelligent transformation of society, especially with the rise of AI models [1] Group 1: Current State of Computing Power in China - As of June 2023, China's total number of operational computing power centers reached 10.85 million standard racks, with intelligent computing power at 788 EFlops, ranking among the top globally [1] - The rapid iteration of AI raises concerns about insufficient computing power [1] Group 2: Huawei's Strategy - Huawei is adopting a differentiated path by constructing a new computing supply system based on "super nodes + clusters" to address the increasing demand for computing power and external limitations in chip manufacturing [2] - The "super node" concept integrates computing chip resources to create a low-latency, high-bandwidth computing entity, enhancing computing efficiency for training and inference of large models [2] - Huawei's latest super node products, Atlas 950 SuperPoD and Atlas 960 SuperPoD, support 8,192 and 15,488 Ascend cards respectively, with the computing power scale of the new clusters exceeding 500,000 and reaching 1 million cards [3] Group 3: Future Outlook and Innovations - Huawei aims to innovate systematically with the "super node + cluster" approach, providing sustainable and scalable computing power for China's AI development [3] - The company plans to advance its computing capabilities at a pace of nearly one generation per year, doubling the computing power each time [3] - Huawei has introduced the UnifiedBus protocol to overcome technical bottlenecks in large-scale super node interconnection, promoting an open ecosystem for computing power [5][6] Group 4: Open Source and Ecosystem Development - Huawei is committed to open-sourcing its Ascend hardware to accelerate developer innovation and build a robust ecosystem [6] - The trend towards open-source technology is seen as essential for the construction of a computing ecosystem, with significant implications for reducing computing costs and fostering comprehensive system innovation [6][7] - Recent government policies are aimed at strengthening the open-source landscape, providing a supportive environment for the development of the computing ecosystem in China [7]

Group 1 - MoE Thread's IPO has been approved, taking only 88 days from application to approval, the shortest time for new companies on the Sci-Tech Innovation Board this year [1] - The company is one of the earliest and fastest domestic GPU companies to commercialize, with AI computing products projected to account for 77.6% and 94.9% of revenue in the first half of 2024 and 2025 respectively [1] - The funds raised will be used for the development of next-generation AI training and inference chips, graphics chips, and AISoC chips, as well as to supplement working capital [1] Group 2 - According to Guotai Junan, this represents a substantial advancement for domestic full-function GPU manufacturers in the capital market, with expectations for accelerated R&D investment and ecosystem development [1] - The focus of competition in the domestic computing power supply side is shifting towards interconnect optimization, with future directions including achieving high bandwidth interconnects at lower costs and higher energy efficiency systems [1] - Jiangsu Securities notes that the semiconductor sector's valuation has significantly recovered due to the AI market, with demand for computing chips driving growth along the supply chain [2] Group 3 - Yingqu Technology holds 134,037.4 shares of MoE Thread, accounting for 0.3351% of its pre-issue total share capital, and is actively developing brain-computer interface technology [3] - Maolai Optics provides optical devices for lithography machines, which are essential for achieving light uniformity and exposure imaging [3]

Core Insights - The next major opportunity in AI lies within "new infrastructure for computing power" as highlighted by the recent measures from the National Development and Reform Commission and other departments [1][3] - The demand for AI large models and intelligent computing centers is expected to drive a high growth cycle in computing power infrastructure, with a significant leap in the industry chain scale projected by Q4 2025 [1][3] Industry Developments - Computing power has become a fundamental resource for societal operations, supporting technologies from facial recognition to autonomous driving [3] - A 1% increase in the computing power index can lead to a 0.33% growth in the digital economy and a 0.18% increase in GDP [3] - Major global tech companies are intensifying investments in computing power infrastructure, with Alibaba planning to invest 380 billion yuan in AI infrastructure over the next three years, marking the highest investment record in China's private sector for AI infrastructure [3][4] - Huawei has launched the world's strongest computing power supernodes and clusters, with Atlas 950 and 960 SuperPoD products leading in key metrics [4][5] - OpenAI, along with SoftBank and Oracle, announced a plan to invest $500 billion over four years to build data centers and supercomputing clusters in the U.S. [5] Market Performance - As of September 26, there are 319 listed companies in the computing power concept sector in A-shares, with 91 companies seeing stock price increases of over 50% this year, and 34 companies achieving a doubling of their stock prices [6][7] - Companies like Pingzhi Information have transitioned to computing power services, signing contracts worth over 2.2 billion yuan for computing power services [8] - Bohui Co., Ltd. has accelerated its transformation, establishing a subsidiary focused on liquid cooling management systems for intelligent computing centers [8][9]