（来源：燕财局） 美国芯片巨头英伟达CEO黄仁勋，再次"口出狂言"。 据美媒Punchbowl News于3月19日报道，黄仁勋在采访中明确表示："等所有美国企业都用上Vera Rubin，我们就该考虑让Blackwell进入中国市场竞争了。" Vera Rubin是英伟达下一代、最先进的AI芯片，计划2026年下半年推出；Blackwell是当前一代AI芯片， 目前被禁止对华出口。 说白了，就是先确保美国本土用上最顶级的芯片，再申请将被淘汰的次品卖给中国，以保住在中国的市 场份额。 1、联合韩国，把中企当冤大头 此前，英伟达已多次为中国市场推出"专供版"芯片（如H100、H800、H20），这些芯片均是在美方要 求下，对性能进行了不同程度的削减。 据过往经验，即便英伟达之后对华出口次级芯片，很有可能会同样进行削减。 而且自2025年10月起，英伟达已和韩国达成合作，向其提供26万颗完整的Blackwell芯片，帮助其构建 国家级的AI算力底座。 3月16日，英伟达还与韩国新世界集团达成合作，投资数十亿美元搭建韩语大模型；在无人驾驶领域亦 有深度合作。 与对华明令禁止的态度不同，美国商务部高度重视并支持对韩合 ...

Core Insights - The article emphasizes the advancements in AI models, particularly focusing on the capabilities of Xiaomi, MiniMax, Anthropic, and Tencent's QClaw in enhancing agent functionalities and their integration into communication platforms [1][2][3]. Group 1: AI Model Advancements - Xiaomi and MiniMax are enhancing their agent capabilities, marking the transition into a new era of executable, collaborative, and self-evolving agents. Xiaomi's MiMo-V2 models are recognized for their strong execution capabilities, with MiMo-V2-Pro supporting 1 million context tokens and offering API pricing at one-fifth of competitors [2]. - MiniMax's M2.7 model has achieved a 97% instruction adherence rate in complex skill environments and is capable of autonomously constructing and optimizing agent harnesses, indicating a significant leap in execution and self-evolution capabilities for domestic AI models [2]. Group 2: Communication Integration - Anthropic and Tencent's QClaw are expanding agent access into instant messaging scenarios. Anthropic's Claude Code has introduced a Channels feature that integrates with Telegram and Discord, allowing users to execute code and receive feedback directly from mobile devices [3]. - Tencent's QClaw has launched public testing with an upgraded WeChat mini-program, enabling file uploads and task execution without the need for coding, showcasing the integration of domestic agent products with high-frequency social platforms [3]. Group 3: Infrastructure Developments - NVIDIA has resumed production of the H200 AI processor for the Chinese market, responding to increased demand and receiving export licenses. The company is also introducing the new MGX NVL rack, which doubles the NVLink domain capacity to accommodate 144 GPUs, indicating a simultaneous advancement in supply restoration and next-generation computing infrastructure [4].

Investment Rating - The report maintains an "Overweight" rating for the computer sector, recommending stocks such as Rilian Technology, Kingsoft Office, Haiguang Information, Inspur Information, Hehe Information, Hikvision, Saiyi Information, New Guodu, Xunce, and Jushuitan [4]. Core Insights - Xiaomi and MiniMax have recently enhanced their Agent capabilities, marking the entry of large models into a strong execution and self-evolution era. The MiMo-V2 and MiMo-V2-Pro support up to 1 million context, achieving first-tier status in Coding Agent and Tool Use dimensions, with API pricing at only one-fifth of competitors [4]. - Anthropic and Tencent's QClaw are expanding their Agent access into instant messaging scenarios, integrating with platforms like Telegram, Discord, and WeChat mini-programs, thus promoting the use of Agents in real-time communication [4]. - NVIDIA has resumed production of the H200 AI processor for the Chinese market and introduced the new MGX NVL rack, which doubles the NVLink domain capacity to accommodate 144 GPUs, indicating a simultaneous evolution of supply and infrastructure in the Chinese market [4]. Summary by Sections Industry Overview - The report highlights the acceleration of the industry with key players like Xiaomi, MiniMax, Anthropic, and Tencent enhancing their model capabilities and entry strategies [2]. Investment Recommendations - The report suggests a focus on companies that are advancing in the Agent capabilities and infrastructure, with specific stock recommendations including Rilian Technology, Kingsoft Office, and others [4]. Technological Developments - The advancements in large models and their applications in various scenarios, such as multi-agent collaboration and office automation, are emphasized, showcasing a significant increase in execution capabilities [4].

Investment Rating - The report maintains an "Overweight" rating for the computer sector, recommending stocks such as Rilian Technology, Kingsoft Office, Haiguang Information, Inspur Information, Hehe Information, Hikvision, Saiyi Information, New Guodu, Xunce, and Jushuitan [4][5]. Core Insights - Xiaomi and MiniMax have recently enhanced their Agent capabilities, marking the entry of large models into a strong execution and self-evolving Agent era. Hunter Alpha and Healer Alpha are confirmed as early versions of MiMo-V2, which supports 1 million context and ranks in the top tier for Coding Agent and Tool Use dimensions, with API pricing at only 1/5 of competitors [4][5]. - Anthropic and Tencent's QClaw have expanded their Agent access to instant messaging scenarios by integrating with Telegram, Discord, and WeChat mini-programs, facilitating easier user interaction with AI [4][5]. - NVIDIA has resumed production of the H200 AI processor for the Chinese market and introduced the new MGX NVL rack, which doubles the NVLink domain capacity to accommodate 144 GPUs, indicating a simultaneous evolution of supply and infrastructure in the Chinese market [4][5]. Summary by Sections Industry Overview - The report discusses the acceleration of model Agent capabilities and entry points, with companies like Xiaomi, MiniMax, Anthropic, and Tencent leading the charge [2][4]. Investment Recommendations - The report suggests a focus on companies that are enhancing their AI capabilities and infrastructure, with specific stock recommendations provided [4][5]. Company Performance Predictions - The report includes earnings per share (EPS) forecasts for recommended companies, indicating growth potential and investment viability [5].

Core Insights - The article discusses the rising importance of "Token" in the tech industry, particularly in the context of AI and its commercialization [1] - The shift from a "model-centric" era to a "Token-centric" era is highlighted, indicating a fundamental change in the business logic of AI [1] Group 1: Token Economy and Market Dynamics - The demand for Tokens has surged, with weekly usage increasing threefold compared to January 2023 [1] - NVIDIA's CEO Jensen Huang projected the market demand for Tokens to rise from $500 billion to $1 trillion by 2027 [2] - Major cloud service providers, including Tencent Cloud and Alibaba Cloud, have announced price increases for AI-related services, with Alibaba Cloud raising prices by up to 34% [2] Group 2: Competitive Landscape - Chinese AI model companies are benefiting from the Token consumption surge, with local models priced significantly lower than OpenAI's offerings, often at 10-20% of the cost [3] - MiniMax's M2.5 model has seen a sixfold increase in daily Token consumption from December 2025 to February 2026, indicating strong market traction [3] - Kimi's revenue has surpassed its entire 2025 earnings within just 20 days, showcasing rapid growth and market acceptance [3] Group 3: Infrastructure and Hidden Winners - AI Data Centers (AIDC) and communication networks are crucial for Token production and transmission, benefiting companies like Runze Technology and Guanghuan New Network [4] - The Token economy represents a restructuring of the value chain in the AI industry, akin to a gold rush where various players can profit [4] - The future of the Token economy will depend on who can produce, deliver, and maximize the value of Tokens most efficiently [4]

Group 1 - Nvidia's CEO Jensen Huang announced the restart and acceleration of H200 production for the Chinese market, indicating a strategic move to increase market share in China [1] - The resumption of H200 production follows the acquisition of export licenses from Chinese clients, marking progress in Nvidia's re-entry into the Chinese market after previous production halts due to regulatory constraints [1] - Several Chinese companies have reportedly received approval to purchase H200 orders, although final confirmation from Beijing remains unclear [1] Group 2 - Nvidia is preparing to launch a Groq chip version for the Chinese market, designed for flexibility and integration with other systems, expected to be available as early as May [2] - The upcoming Vera Rubin GPU will be used for model training in overseas data centers, while the Groq chip will handle inference operations, with both products deployed separately to comply with export control regulations [2]

Market Overview - Stocks experienced a pullback due to rising oil prices and higher-than-expected Producer Price Index (PPI) data, with Brent Oil surpassing $108 and WTI around $98 [2][3] - The PPI increased by 7%, with core PPI rising by 0.5%, not accounting for energy prices, indicating inflation risks are on the rise [3][4] Federal Reserve Insights - The Federal Reserve is expected to maintain its current stance during the upcoming meeting, with a focus on communication regarding inflation and economic projections [3][4] - The dot plots from the Fed will provide insights into expectations for inflation and GDP through the end of the year and into next year [5][6] Energy Sector Analysis - Energy is viewed as a key hedge in the current market environment, with investors encouraged to increase exposure to this sector due to its resilience [10][11] - Only utilities and energy sectors have over 50% of stocks above their 200-day moving averages, indicating strength amidst broader market volatility [12] Technology Sector Dynamics - The technology sector shows mixed performance, with notable strength in software and memory industries, particularly with Micron expected to report significant revenue growth [17][18] - Micron's anticipated revenue of $19.77 billion represents a 146% increase from the previous year, highlighting strong demand in the memory market [18][19] Investment Strategy - Investors are advised to be selective within the technology sector, moving beyond major tech stocks to identify pockets of strength [20][21] - The current market conditions favor stock pickers, allowing for potential gains in specific sectors despite overall market volatility [21][22]

关于向中国出口人工智能（AI）半导体，黄仁勋3月17日表示"已经获得了面向众多客户的许 可"，并透露"面向中国的生产已在重启"…… 美国商务部长卢特尼克今年1月在接受CNBC采访时表示，目标是将台湾约四成的半导体生产 转移到美国。对于这一目标的可行性，黄仁勋回应称"非常困难"。他还表示，尽管其委托生 产的台积电正在扩大对美投资，但对台湾的依赖仍将持续。 美国半导体巨头英伟达首席执行官黄仁勋3月17日在正在举行的技术活动GTC上召开记者会 表示，关于向中国出口人工智能（AI）半导体，"已经获得了面向众多客户的许可"，并透 露"面向中国的生产已在重启"。 版权声明：日本经济新闻社版权所有，未经授权不得转载或部分复制，违者必究。 日经中文网 https://cn.nikkei.com 黄仁勋表示，对华出口"与两周前相比情况已经发生变化"，并称"已经接到许多客户的订 单"。特朗普领导的美国政府曾在2025年批准英伟达高性能半导体"H200"对华出口，但许可 范围仍然较为有限。 黄仁勋指出，"特朗普总统的意图是让美国在获取最先进技术方面占据主导地位"。英伟达此 前也向美国政府反映，如果对华出口管制持续，可能反而会助长中 ...

Summary of the Conference Call on China's AI GPU Outlook and TSMC's Capital Expenditure Industry Overview - The conference focused on the outlook for China's AI GPU market and the latest capital expenditure expectations from TSMC, highlighting the increasing importance of domestic chip production in the AI sector [1][2][3]. Key Points and Arguments 1. **AI Chip Ownership**: Companies like Alibaba are moving towards owning their chips to reduce reliance on third-party suppliers, similar to Google's strategy with TPU [5][6]. 2. **Customization and Flexibility**: Owning chips allows companies to tailor their products to specific applications and adjust capacity based on demand, which is crucial in the rapidly evolving AI landscape [7][8]. 3. **Performance Comparison**: Domestic chips are reportedly closing the performance gap with international counterparts, with some Chinese chips outperforming NVIDIA's A100 in inference tasks [9][10]. 4. **Market Positioning**: Alibaba is positioned as a leading player in the AI space due to its comprehensive supply chain, including its chip production (Pingtouge) and cloud services [12][13]. 5. **Demand Projections**: The demand for AI chips in China is expected to grow significantly, with projections estimating a market size of $67 billion by 2030, driven primarily by internet companies [14][15]. 6. **Supply Chain Dynamics**: The supply of AI chips is anticipated to increase, with domestic foundries like SMIC and Hua Hong playing key roles in supporting production [27][28]. 7. **Self-Sufficiency Goals**: The self-sufficiency rate of domestic AI chips is projected to rise from 33% in 2024 to 76% by 2030, indicating a strong push towards local production [27][28]. 8. **Valuation Insights**: Valuations for companies like Kunlun and Pingtouge were discussed, with estimates suggesting a market cap range of $20 billion to $61 billion for Kunlun based on a price-to-sales ratio of 26x [19][20]. Additional Important Insights - **Investment Recommendations**: Alibaba was highlighted as a preferred investment due to its strong position in the AI ecosystem and expected performance in upcoming earnings reports [21][22]. - **Global Competitive Landscape**: The conference noted that while domestic companies are gaining ground, competition remains fierce, particularly from state-owned enterprises like Huawei and Cambricon [16][17]. - **Technological Advancements**: The discussion included the importance of advancements in chip technology and packaging, with Chinese firms catching up in areas like 2.5D packaging and advanced process nodes [29][30]. - **Market Consolidation**: The AI chip market is expected to undergo consolidation, with a few key players dominating the landscape, which may lead to reduced margins for new entrants [17][37]. This summary encapsulates the critical insights from the conference call, emphasizing the strategic shifts in China's AI chip industry and the implications for investment and market dynamics.

Investment Rating - The report does not explicitly provide an investment rating for the industry Core Insights - The AI industry is experiencing strong demand, with companies like Nvidia and Marvell reporting significant revenue growth due to AI applications [8][19][20] - Nvidia's strategic investments in silicon photonics technology aim to alleviate supply chain bottlenecks and support the scaling of AI infrastructure [18][31] - OpenAI and Anthropic have achieved substantial annual revenues, indicating a robust market for AI models [14][15] Industry Frontiers - Samsung plans to continue raising NAND product prices in Q2, following a 100% increase in Q1, indicating ongoing supply constraints in the semiconductor market [12] - Nvidia has ceased production of the H200 AI chip for the Chinese market, reallocating resources to the next-generation AI computing platform [13] - OpenAI's GPT-5.4 has demonstrated advanced capabilities, surpassing human-level performance in certain tasks [14] Capital Trends - Nvidia's $20 billion investments in Coherent and Lumentum highlight a commitment to developing next-generation silicon photonics technology [18] - Companies like Broadcom, Marvell, and Ciena have reported strong financial results, reflecting the increasing demand for AI-related products [19][20][21] - ByteDance's partnership with Century Internet for a 500 MW data center underscores the growing need for AI computing power in China [23] Weekly Perspective - The report emphasizes the strong performance of companies in the AI sector, with Broadcom and Marvell both reporting significant revenue increases [31] - Ciena's focus on AI training and co-packaged optics technology is seen as a response to the limitations of traditional electrical interconnects [31] - MiniMax's impressive revenue growth indicates a successful commercialization of AI technologies, driven by improved model efficiency [31]