上周最新季度财报发布后的电话会议上，不仅英伟达CEO黄仁勋回应了AI泡沫论，表示英伟达看到 了不一样的东西，且现在已进入AI良性循环，英伟达CFO科莱特·克雷斯（Colette Kress）也驳斥 了英伟达芯片使用寿命不长的说法，称6年前的芯片仍在满负荷工作。 2025.11. 25 本文字数：1361，阅读时长大约3分钟 作者 | 第一财经 郑栩彤 AI泡沫相关的议论不断，英伟达也下场回应了相关的质疑。 相关的争论仍未停止。曾在2008年次贷危机前做空房地产的"大空头"迈克尔·伯里（Michael Burry）近日在社交媒体上发表了关于英伟达的系列言论，关于科技公司围绕英伟达的循环交易和投 资者收益减少等提出批评，引来英伟达反击。 迈克尔·伯里在社交平台上发布了一张图，描绘了以英伟达为中心的美国科技公司投资、采购关系 图，包括甲骨文花费数百亿美元采购英伟达芯片、OpenAI与甲骨文之间有3000亿美元云相关的交 易、英伟达计划至多投资OpenAI1000亿美元，图中写出的公司还包括英特尔、AMD、xAI、微软 等。 "以下列出的每家公司都存在可疑的收入确认方式。若将所有交易往来的情况以图表形式呈现，那会 是 ...

英伟达也在备忘录中反驳了这一观点，称2018年以来英伟达已回购价值910亿美元的股票，而非1125亿 美元，迈克尔·伯里似乎错误地将 RSU（限制性股票单位）的税款计算在内了。英伟达还指出，员工的 股权授予不应与回购计划执行情况混为一谈。英伟达的员工薪酬与同行水平一致，员工因股价上涨而受 益并不意味着授予股权过于慷慨。 迈克尔·伯里并不认同英伟达的反击，随后他表示，英伟达试图在备忘录中反驳他的观点，但他仍坚持 自己的分析观点，他还将发布相关的其他内容。 迈克尔·伯里在社交平台上发布了一张图，描绘了以英伟达为中心的美国科技公司投资、采购关系图， 包括甲骨文花费数百亿美元采购英伟达芯片、OpenAI与甲骨文之间有3000亿美元云相关的交易、英伟 达计划至多投资OpenAI1000亿美元，图中写出的公司还包括英特尔、AMD、xAI、微软等。 "以下列出的每家公司都存在可疑的收入确认方式。若将所有交易往来的情况以图表形式呈现，那会是 一幅难以理解的复杂图景。未来人们会将此视为欺诈行为的证据，而非一个良性循环的模式。"迈克尔· 伯里表示，真正的最终需求量极其微小，几乎所有的客户都是由其经销商提供资金支持的。 对此，英伟 ...

国产 AI 芯片厂商竞争格局、产品力与市场情况 20251123 摘要 国产 AI 芯片在部分模型性能上可达英伟达 A100 的 50%-100%，推理 侧华为、寒武纪、昆仑芯达 H100 的 50%-70%，但尚无芯片达到 H100 水平。受美国限制，国产芯片依赖海外产能，少数如华为可获国 内 7 纳米（N+2）产能，多数厂商仅能获得 12 纳米（N+1）制程。 国内 AI 芯片厂商计划在 2026 年 Q1-Q3 量产新一代全国产 12 纳米 （N+1）制程芯片，华为、寒武纪、海光预计上半年实现 7 纳米 （N+2）制程量产，但这些产品多为上一代加强版。华为和寒武纪已开 发出基于光电直连的新型互联方式，部分超越英伟达 NVLink 技术。 二三线 AI 芯片设计公司采用迷你超节点技术，通过定制高速网卡连接 32-64 张 GPU 卡以提高性能。百度昆仑已完成万卡集群调试，并开发 类似平头哥超节点系统，通过与上下游合作提升竞争力。 2025 年 Smith N+2 芯片月产量 4,000-5,000 片晶圆，良率接近 40%，预计 2026 年产量增至每月超 1 万片，但不超 1.5 万片，良率上 限 50 ...

H200放开的消息今天已经传的沸沸扬扬了，国内的新闻基本都是这样写的： 但这个新闻最早 是出自彭博，比路 透要早2个多小时。 而彭博的新闻是下面这个写的，也就是说根据彭博的这个描述，目前只是初步讨论，而且完全有可 能只是停留在讨论，永远不会放开。 这事还得回溯到前段时间中美领导层见面，川普说会谈到Blackwell，大家都以为B30A会放开。后来 的事大家也都知道了，川普说没有谈Blackwell。 但又过了两天，WSJ上的消息说是因为川普的高级顾问们都反对，所以才没有谈，我们当时在星球 中就发过这个： 两国领导开会那天上午，有朋友就发我这样的截图： 所以可能高端的Hopper要放开的事也讨论了很久了。 说话正题，这次的说法是H200要放开，先看下H200的性能： | Specification | H100 | H200 | | --- | --- | --- | | GPU Architecture | Hopper | Hopper | | GPU Memory | 80 GB HBM3 | 141 GB HBM3e | | GPU Memory Bandwidth | 3.35 TB/s | 4.8 ...

Group 1: Economic Indicators - U.S. manufacturing activity has contracted for the eighth consecutive month in October, with the ISM manufacturing index declining by 0.4 to 48.7 [12] - The U.S. Treasury estimates borrowing of $569 billion in the fourth quarter, a decrease of $21 billion from the July forecast [12] - China's manufacturing PMI fell to 50.6 in October, indicating a slowdown in manufacturing expansion [12] Group 2: Central Bank Policies - Federal Reserve officials suggest a potential interest rate cut in December, contingent on forthcoming economic data [12] - The People's Bank of China has renewed a bilateral currency swap agreement with the Bank of Korea [11] Group 3: Commodity Markets - WTI crude oil closed at $60.82 per barrel, up 0.26%, while Brent crude oil rose 0.27% to $64.65 per barrel [7] - Gold prices fluctuated around $4000 per ounce, closing down 0.07% at $4000.44 [7] Group 4: Stock Market Performance - U.S. stock indices showed mixed results, with the Dow Jones down 0.48% and the S&P 500 up 0.17% [5] - Hong Kong's Hang Seng Index rose 0.97%, with significant trading volume of 228.68 billion HKD [5] - A-shares saw the Shanghai Composite Index increase by 0.55% [6]

Group 1: Market Overview - US stock indices opened higher, with Dow Jones up 0.06%, S&P 500 up 0.49%, and Nasdaq up 0.97% [1] - Amazon surged over 4% after signing a $38 billion agreement with OpenAI to expand computing power using Nvidia chips [1] - Nvidia rose over 2% as Microsoft received export licenses to ship Nvidia's A100, H100, and H200 GPUs to the UAE [1] - Google opened slightly higher by 0.1%, planning to issue $15 billion in dollar bonds [1] Group 2: Company News - Microsoft plans to invest nearly $8 billion in the UAE by 2029, with over $5.5 billion allocated for AI and cloud infrastructure expansion [2] - IREN announced a $9.7 billion GPU cloud service contract with Microsoft, including a 20% upfront payment, and a separate $5.8 billion agreement with Dell for GPU and related equipment [3] - BP announced the sale of non-controlling interests in US onshore assets for $1.5 billion, reducing its stake in certain areas [4] - Kimberly plans to acquire Kenvue for approximately $48.7 billion through a cash and stock transaction [5] - BASF will integrate its Asian PolyTHF business into its China facility and close its Ulsan plant in South Korea by 2026 [6]

Core Viewpoint - The article discusses the rise of Cambrian, a company specializing in AI chips, which has recently surpassed Moutai to become the "king" of A-shares, highlighting its rapid growth and technological advancements in the AI chip sector [2][3]. Group 1: Company Background - Cambrian was founded in 2016 by two talented brothers from Nanchang, Jiangxi, focusing on AI chip development [4][6]. - The brothers, Chen Yunji and Chen Tian Shi, have impressive academic backgrounds, with both achieving significant milestones in their education and research [6][9]. - The idea to develop AI-specific chips emerged in 2010, during a time when AI technology was not widely recognized [9]. Group 2: Technological Development - Cambrian developed the world's first deep learning-specific processor prototype within five years, showcasing its commitment to innovation [10]. - The company gained significant recognition in 2017 when Huawei's AI phone chip, Kirin 970, utilized Cambrian's 1A processor, marking a pivotal moment in its growth [16]. Group 3: Financial Performance - Cambrian's valuation skyrocketed from $100 million in 2017 to over $1 billion in 2018, and by July 2020, its market capitalization exceeded 100 billion RMB upon listing on the STAR Market [17]. - Despite facing a prolonged downturn post-IPO due to competition and internal challenges, Cambrian decided to focus on independent chip product development, leading to substantial net losses over four consecutive years [21]. - The company finally achieved profitability in Q4 of the previous year, with a staggering 43-fold increase in revenue in the first half of this year compared to the same period last year [23]. Group 4: Recent Developments - The launch of the Si Yuan 590 chip in 2024, which is comparable in performance to NVIDIA's A100, positions Cambrian favorably in the domestic AI chip market [22]. - Cambrian's financial report for the first half of the year shows significant growth, with total revenue reaching approximately 2.88 billion RMB, a 4,347.82% increase year-on-year [24]. - The company's net profit attributable to shareholders was over 1 billion RMB, marking a significant turnaround from previous losses [24].

Summary of Conference Call Records Company and Industry Overview - **Company**: Haiguang Information - **Industry**: AI Chip Market and Cloud Service Providers (CSP) Key Points and Arguments Capital Expenditure and Market Trends - Alibaba expects its capital expenditure for Q2 to remain flat compared to Q1, with an annual CSP capital expenditure projected to maintain around 100 billion, reflecting ongoing investment in computing infrastructure [2][3] - Major domestic CSPs are accelerating capital expenditures in the second half of the year, with Tencent's Q1 expenditure at 27.5 billion and Q2 at 19.1 billion, slightly below expectations [3] AI Chip Market Development - The domestic AI chip market is transitioning from bidding to domestic substitution, with companies like Cambricon increasing chip orders, while Nvidia faces delays in chip deliveries due to foreign trade issues [2][4] - Haiguang Information reported a revenue of 5.464 billion in H1 2025, a 45.21% year-on-year increase, with a net profit of approximately 1.2 billion, up 41% [4][16] Future Demand for AI Computing Power - Future AI computing demand is expected to arise from three main areas: AI transformation in traditional industries, large model applications in internet companies, and B-end projects from state-owned enterprises [6] - Traditional industry chip demand is projected to grow 2-3 times, while internet companies like ByteDance expect a tenfold increase in TOKEN sales [6] Performance and Capacity of AI Chip Manufacturers - There is a significant performance and capacity gap between domestic and international AI chip manufacturers, with Nvidia's GPU200 outperforming domestic chips like A100/A7,100 by several times [7] - Domestic chips are competitively priced but lag in performance, with major domestic players like Huawei and Cambricon making strides in the market [7][13] Industry Trends and Future Outlook - The AI computing industry is expected to accelerate domestic chip substitution, expand application scenarios, and enhance overall performance [8] - High-performance cards like Huawei's Ascend 384 cluster may face supply shortages, while independent high-performance cards from Cambricon are set to secure production capacity in H1 2025 [8] Financial Performance and Projections for Haiguang Information - Haiguang Information's H1 2025 revenue is projected at 9 billion, with GPU revenue contributing 2 billion, and a net profit forecast of 4 billion, corresponding to a market value of over 400 billion [19] - The company’s costs increased by 58% year-on-year, primarily due to a higher proportion of new products and increased material costs [17] - Inventory levels reached 6 billion, with contract liabilities at 3.091 billion, indicating strong market demand [18] Competitor Performance - Other companies like Inspur and Dawning also reported significant revenue growth, with Inspur achieving approximately 80 billion in H1 2025, a 90% increase year-on-year [21] - Dawning's performance remained stable, with a slight increase in revenue and net profit [22][23] Additional Important Insights - The AI server market is experiencing stable and rapid growth, with companies like Inspur and Dawning being noteworthy players [15] - The overall market sentiment indicates a high level of optimism regarding the future of AI chips and computing power, driven by increasing demand across various sectors [8][19]

Core Viewpoint - The recent surge in Cambricon's stock price, reaching a peak of 1464.98 yuan, highlights the excitement in the AI chip market, raising questions about whether this is a golden era for AI chips or a capital bubble [1] Group 1: Cambricon's Surge - Cambricon's stock price increase is driven by three main factors: explosive revenue growth, with a projected revenue of 2.881 billion yuan in the first half of 2025, a year-on-year increase of 4347%, and a net profit of 1.038 billion yuan, marking a turnaround to profitability [3] - The recent policy support from the government, particularly the "Artificial Intelligence +" action plan aiming for over 70% AI penetration by 2027, positions Cambricon favorably as a rare listed AI chip company [3] - The challenges faced by Nvidia, including safety risks associated with its H20 chip, have led to increased expectations for Cambricon's orders as domestic chips become a "safety choice" for government and enterprise markets [3] Group 2: Valuation Concerns - Historical data shows that 15 companies that surpassed Moutai's market cap often fell back due to valuation bubbles, with Cambricon's dynamic P/E ratio at 343 times and P/S ratio at 111 times, compared to Intel's 4.2 times [4] - The competitive landscape for Cambricon is intense, with rivals like Huawei and Baidu rapidly advancing, indicating that its technological moat may not be as robust as perceived [4] Group 3: Risks and Challenges - Cambricon faces customer concentration risk, with its largest client accounting for 79% of revenue in 2024, which is expected to decrease to 42% in 2025, emphasizing the need for diversification in its client base [5] - The company's reliance on SMIC's 7nm process technology poses a risk if further U.S. sanctions occur, potentially hindering its ability to innovate and compete effectively [5] - The sustainability of policy support for domestic chip alternatives is uncertain, and any reduction in local government subsidies could lead to a sharp decline in performance [6] Group 4: Long-term Outlook - In the short term, Cambricon may benefit from market sentiment and policy support, but for long-term success, it must achieve significant technological breakthroughs, diversify its customer base, and ensure stable profitability [7]

Core Viewpoint - Cambricon, known as the "first domestic AI chip stock," reported a significant revenue increase of 4347.82% year-on-year for the first half of 2025, reaching 2.881 billion yuan, and achieved a net profit of 1.038 billion yuan, marking a turnaround from losses to profitability [2][3]. Financial Performance - In the second quarter of 2025, Cambricon's revenue was 1.769 billion yuan, a quarter-on-quarter increase of 59.19%, with a net profit of 683 million yuan, up 92.03% quarter-on-quarter [3]. - The cloud product line accounted for 99.62% of total revenue, amounting to 2.870 billion yuan [3]. Market Reaction - Following the release of the half-year report, Cambricon's stock price surged over 7% at the market opening on August 27, closing up 6.01% at 1408.9 yuan per share, with a market capitalization nearing 600 billion yuan [2]. - The price-to-earnings ratio (TTM) adjusted from over 4000 times to around 500 times, still significantly higher than competitors like Nvidia and AMD [2]. Capital Raising and Future Plans - Cambricon's recent 4 billion yuan targeted placement plan has been approved by the Shanghai Stock Exchange, awaiting final registration from the China Securities Regulatory Commission [4]. - The funds will primarily support projects related to chip platforms for large models and software platforms to enhance chip usability and adaptability [4]. Competitive Landscape - The domestic chip replacement trend is ongoing, with local governments setting targets for the use of domestic chips in new computing centers [5]. - Cambricon faces competition from industry giants like Nvidia, which is developing new AI chips, and has recently been involved in security controversies regarding its products [5]. Market Sentiment and Speculation - Speculation about Cambricon's order volume and production capacity has fueled stock price increases, although the company has denied rumors of significant orders from major clients [5][6]. - The release of DeepSeek's new model, which supports the next generation of domestic chips, has also contributed to positive market sentiment [7]. Industry Developments - The introduction of FP8 precision formats by DeepSeek is expected to enhance compatibility and performance for domestic chips, with several manufacturers claiming support for this new standard [10][11]. - The use of FP8 precision is seen as crucial for training large models efficiently, and its adoption is anticipated to grow among AI chip manufacturers [10][11].