格隆汇11月26日｜富士康：将扩大在美国威斯康星州人工智能基础设施方面的投资。威斯康星州经济发 展公司已批准富士康追加投资5.69亿美元，以扩大其在拉辛县的业务运营。 ...

以做空次贷闻名的"大空头"投资者迈克尔·伯里（Michael Burry），一直是当前AI热潮的直言不讳的质 疑者。在关闭对冲基金数日后，他推出了一个聚焦金融市场的付费Substack通讯，并将矛头对准了英伟 达。 "当今AI热潮的五大公开巨头——微软、谷歌、元宇宙、亚马逊和甲骨文，再加上几家新兴初创公司， 承诺在未来三年内投入近3万亿美元用于人工智能基础设施。投资者对此简直爱不释手，"他写道，"而 这一切的核心，又是一个当年思科式的角色——为所有人提供'铲子和镐头'，并怀揣着宏大愿景。它的 名字叫英伟达。" 注：思科是互联网泡沫破灭最具标志性的"坟场级"案例。在互联网泡沫时期，思科的股价因预期中的互 联网流量而飙升，但当预期未能立即兑现时，股价便遭受重创。 伯里近期加大了对科技巨头的批评力度，包括英伟达和帕兰提尔科技公司，他质疑云基础设施热潮，并 指责主要供应商利用激进的会计手段，夸大其巨额硬件投资带来的利润。 该通讯目前已有超过2.1万名订阅者，月费39美元。他暗示发布频率通常是每周至少一篇，有时会更 多。 在题为《基础：我的1999年（及2000年部分时期）》的文章中，伯里讲述了自己早期的兴趣以及如何开 ...

Core Insights - The world's largest technology companies are raising nearly $100 billion through recent bond offerings to enhance their artificial intelligence infrastructure [1] Group 1 - Major technology firms are actively tapping into debt markets [1] - The total amount raised through these bond offerings is close to $100 billion [1] - The primary motivation for these bond issuances is to bolster AI infrastructure [1]

国泰君安期货研究所· 海 外 研 究 戴璐 Z0021475 国泰君安期货·君研海外 基本面、降息预期和"AI泡沫" ——透视当前海外三大风险点 资料来源：Bloomberg，国泰君安期货研究 Special report on Guotai Junan Futures (本报告感谢杨藤贡献） 日期：2025年11月24 Guotai Junan Futures all rights reserved, please do not reprint 三大逻辑点落地，市场如何定价？ 11月以来，全球市场经历较多地缘、宏观和产业逻辑的冲击，主要大类资产表现整体呈现几个特征：1）科技板块引领Risk-on，权益 转向防御；2）缺乏明显有效的避险资产，贵金属高位回落，全球国债市场表现不佳；3）年底市场微观流动性不佳，上风险情绪有限 非官方就业指标指向了本次非农数据改善，制造业PMI、服务业PMI、中小 企业就业和CB就业分项集合指数有改善 当前海外市场的波动源自三大逻辑点：1）经济基本面如何？2）基于基本面，对于短期（12月）和远期（2026年）的降息预期如 何？3）从"AI 泡沫"这一更务虚的叙事，转为对AI投资持续 ...

Group 1: Currency and Investment Trends - Offshore RMB (CNH) has the lowest adjusted 3-month carry yield among major Asian currencies at approximately -0.20, making it less attractive for capital inflows aimed at arbitrage compared to higher-yield currencies [6] - Morgan Stanley indicates that this low yield supports the view that the offshore RMB to USD exchange rate will continue to fluctuate within a range [6] Group 2: Real Estate Market Insights - UBS forecasts that the national residential inventory digestion cycle will extend from approximately 30 months to 25 months, with stabilization of housing prices expected by mid-2027, delayed from early 2026 [8] - The forecast faces multiple downside risks, including weaker-than-expected sales recovery and rising mortgage default rates, which could exert new downward pressure on housing prices [8] Group 3: International Travel and Economic Impact - Japan is projected to be the largest international air travel destination for Chinese tourists by 2025, with an estimated 66,150 flights planned, significantly higher than South Korea and Thailand [11] - Nomura estimates that restrictions on Chinese tourists traveling to Japan could result in a GDP loss of up to 0.36 percentage points for Japan [11] Group 4: Manufacturing and Investment Shifts - Chinese companies are accelerating the relocation of manufacturing facilities to Southeast Asia, with the region's share of China's total outbound direct investment rising from nearly 9% in 2017 to close to 18% by 2024 [13] - China's reliance on high-end capital goods from Germany is decreasing, particularly in automotive parts and machinery, as domestic manufacturing capabilities improve [16] Group 5: Automotive Market Developments - The UAE has overtaken Russia to become China's second-largest export market for passenger vehicles, with exports reaching 367,800 units in the first nine months of 2025, surpassing 357,700 units to Russia [18] - Chinese automotive brands are rapidly increasing their market share in the UAE, rising from 10%-14% last year to 15%-20% in the first nine months of this year [20] Group 6: Economic Sentiment and Consumer Behavior - In the U.S., the disparity in economic confidence between high-income and low-income groups is widening, with high-income households maintaining a growth expectation of 3.0%-3.5%, while low-income households' expectations have dropped to around 2.0% [31] - Food price inflation disproportionately affects low-income families, with the lowest income decile spending 45% of their after-tax income on food, compared to about 15% for the second-lowest income group [28] Group 7: Corporate Financing and AI Investments - Amazon plans to issue approximately $12 billion in bonds to finance its AI infrastructure, marking its first dollar bond issuance in three years, as part of a broader capital race in the tech sector [37] - OpenAI and Anthropic exhibit contrasting strategies in capital utilization and profitability paths, with OpenAI expected to incur significantly higher cash consumption before achieving profitability by 2030 [42] Group 8: European Economic Outlook - The European Commission has raised its GDP growth forecast for the Eurozone in 2025 to 1.3%, up from 0.9%, due to stronger-than-expected economic performance in the first nine months of the year [47] - European companies are generally lowering their earnings expectations for the 2025 fiscal year, with 19 out of 22 sectors experiencing downward revisions, particularly in cyclical industries like paper and packaging, and automotive [50] Group 9: Commodity and Energy Market Dynamics - Russian Urals crude oil prices have sharply declined as sanctions against major Russian energy companies come into effect, with the price gap between Urals and Brent crude widening to $23.52 per barrel [65]

信息在亚特兰大微软Fairwater站点的所有GPU之间共享。 高耗能的人工智能基础设施建设持续推进，AI基建竞赛再度升温。微软12日也宣布，将构建首个人工 智能超级工厂。微软正在佐治亚州亚特兰大建设一个新的数据中心，并表示该中心将与威斯康星州另一 座数据中心相连，组成一台"大型超级计算机"，搭载数十万块英伟达芯片，为人工智能技术提供支撑。 这些被称为Fairwater的微软AI数据中心彼此直接相连，最终还将与美国各地在建的其他数据中心互联互 通，通过一种新型专属网络实现数据高速流转。分布在不同州的Fairwater站点能够协同运作，形成AI超 级工厂，加快新一代AI模型训练，将原本需要数月的任务缩短至数周内。 对标OpenAI 1.4万亿美元基础设施建设计划，由亚马逊支持的美国AI初创公司Anthropic拟斥资500亿美 元建造美国人工智能基础设施，率先启动得克萨斯州、纽约州定制化数据中心的建设，首批设施将于 2026年投入运营。 当地时间11月12日，Anthropic宣布与英国AI云平台企业Fluidstack合作开发AI基础设施，支撑业务快速 增长及长期研究规划。由于Fluidstack建设速度快， ...

Core Insights - Amazon has signed a $38 billion agreement with OpenAI, reigniting interest in artificial intelligence and driving up stock prices on Wall Street and among major tech companies [1] - The spending surge in the AI sector continues, with companies like Amazon, Microsoft, and Alphabet investing billions in GPUs and AI infrastructure [1] - Despite concerns over the rapid increase in AI capital expenditures, investors remain engaged, with tech giants leading the market rally [1] Company Developments - Amazon's partnership with OpenAI is a significant move, reflecting its commitment to advancing AI technologies [1] - Major tech companies are collectively investing tens of billions in AI-related resources, indicating a robust growth trajectory for the sector [1] Market Reactions - The agreement has positively impacted stock markets, with Wall Street and large tech stocks experiencing gains [1] - Asian stock markets are expected to rise, influenced by the developments in the AI sector and the performance of tech stocks [1] - Japanese and South Korean stock index futures are showing upward trends, while the Australian stock market remains stable with unchanged interest rates [1]

Group 1: AI Infrastructure Investment in India - Major US tech companies, including Google, Microsoft, and OpenAI, are committing over $10 billion to build AI infrastructure in India, attracting investor interest in local companies linked to data center construction [1][5] - The investment in AI infrastructure is expected to drive the development of auxiliary businesses related to data centers, including core power equipment manufacturers and large power generation companies, with the Indian data center market projected to exceed $100 billion by 2027 [1][2] Group 2: Investment Opportunities in Indian Stock Market - The creation of an AI ecosystem in India offers significant investment opportunities in the $5.4 trillion Indian stock market, which has lagged behind global markets due to a lack of pure AI hardware companies [2] - Investors are shifting focus to local data center auxiliary companies, which are expected to benefit substantially from the infrastructure development for AI data centers, making it one of the hottest equity trades globally [2] Group 3: Key Players and Their Investments - Google plans to invest approximately $15 billion in AI infrastructure in southern India, while Microsoft has announced an initial investment of around $3 billion to expand its cloud computing and AI capabilities [5] - AdaniConneX, a joint venture involving Adani Enterprises and Bharti Airtel, is collaborating with Google on a large AI project, while Reliance Industries is expected to invest up to $15 billion in a one-gigawatt AI data center [6] Group 4: Electrical and Core Power Equipment - A significant portion of capital expenditure related to AI data centers is expected to flow into electrical and core power equipment, including manufacturers of core switchgear, transformers, and distribution systems [8] - Notable Indian stocks in this sector include Hitachi Energy India Ltd., Siemens Ltd., Schneider Electric Infrastructure Ltd., and ABB India Ltd., with the power and related costs accounting for about 40% of data center capital expenditure [8] Group 5: Liquid Cooling Solutions - The high energy consumption of AI workloads generates substantial heat, necessitating advanced liquid cooling systems, with Blue Star Ltd. and Voltas Ltd. being major local providers of such solutions [9] Group 6: Servers and Cloud Infrastructure - Large-scale AI data centers require extensive physical computing infrastructure, with Netweb Technologies India Ltd. recognized as a key player in servers, storage, and AI supercomputing solutions [10] - E2E Networks Ltd. focuses on providing cloud-based AI computing infrastructure and high-performance computing solutions [10]

Group 1 - The core point of the articles highlights the record electricity consumption in the U.S. driven by the expansion of artificial intelligence and data centers, with nuclear energy becoming a focal point to meet this demand [1][2] - Google and NextEra Energy plan to restart the Duane Arnold Energy Center in Iowa by 2029 to provide carbon-free power for Google's AI operations, following its closure in 2020 due to high operational costs and safety concerns [1] - The U.S. electricity consumption is projected to reach 4.097 trillion kilowatt-hours in 2024, with further increases expected, potentially reaching 4.191 trillion kilowatt-hours in 2025 and 4.305 trillion kilowatt-hours in 2026, primarily driven by AI data centers [2] Group 2 - Major tech companies like Oracle, Microsoft, and Google are exploring nuclear energy solutions to address the increasing energy demands from AI computing, with Oracle designing a data center powered by three small nuclear reactors [2] - In addition to the Duane Arnold Energy Center, other nuclear projects are in progress, although no decommissioned nuclear plants have successfully restarted yet [3]

Core Insights - OpenAI emphasizes the need for the U.S. to significantly increase investments in renewable energy capacity to maintain a competitive edge over China in the artificial intelligence development race [1][2] Group 1: Strategic Importance of Energy - OpenAI states that electricity is not just a utility but a strategic asset crucial for building AI infrastructure, which is essential for maintaining leadership in key technological fields [2] - The company encourages the U.S. to commit to constructing 100 gigawatts of renewable energy capacity annually, equating to the annual electricity consumption of approximately 8 million American households [2] Group 2: Competitive Landscape - OpenAI highlights that China's new power generation capacity added last year significantly outpaced that of the U.S., creating an "electricity gap" that poses a risk of the U.S. falling behind [2] - The assertion that "electricity is the new oil" underscores the critical role of energy in the future technological landscape [2]