格隆汇12月5日｜美国银行策略团队提出"空窗期"理论，警告称AI资本支出增长与收入转换的时间差恐 引发市场短期波动。美银美股与量化策略主管 Savita Subramanian指出，虽然超大规模科技公司将资本 支出占比从10年前30%提升至60%，仍远低于互联网泡沫时期140%峰值，但微软、亚马逊等巨头的年度 投入将从今年的4000亿美元跃升至2026年的5100亿美元。对比历史周期，美银指出当前市场存在本质差 异，股票配置比例显著低于泡沫时期，获利成长切实支撑估值体系，IPO规模可控且对未获利企业的投 机热情明显消退。 ...

此外，泡沫破灭的积极意义，还在于为技术"民主化"和应用大爆发扫清障碍。互联网泡沫后，宽带普 及、开源软件兴起、开发工具成本骤降，创业门槛大幅降低，直接催生了Web 2.0黄金时代，诞生了 Facebook、YouTube等深度融入并重塑生活的应用。AI的发展也势必循此路径前行。当前大模型训练与 推理成本高昂，技术门槛将中小玩家挡在门外。泡沫破裂将倒逼技术提供商优化成本、推出更易用廉价 的API服务，让AI技术从实验室和巨头军备竞赛中"解放"，真正赋能千行百业。 韩和元 1995年至2001年间，全球资本市场曾上演过这样一幕：过度投机，估值远超实际价值。最终泡沫破裂导 致市场暴跌、大量企业倒闭。金融史将其称为"互联网泡沫"。如今，AI（人工智能）领域似乎正上演相 似剧情：估值高企、资本扎堆、概念炒作泛滥，种种迹象似乎印证着AI泡沫的存在。 但我想说的是，纵或AI泡沫客观存在，亦大可不必恐慌。诚如互联网泡沫所揭示的：泡沫破灭后，接 踵而至的必然是一场惨烈的出清，但这场出清并未摧毁互联网本身，而是淘汰了缺乏坚实商业模式的投 机者——亚马逊股价暴跌90%后凭借客户中心与物流护城河浴火重生，而单纯讲故事、烧钱换流量的公 ...

也正是基于2000年互联网泡沫的经验和教训，我们不应因此而否定AI的变革潜力，甚至因噎废食地放 弃。恰恰相反，历史的启示告诉我们，当泡沫退去、价格回归理性之日，正是我们理应鼓起勇气，进一 步加大战略性投入之时。面对AI泡沫，恐慌与放弃是下策，成为"理性乐观主义者"才是明智选择。 对于政策制定者而言，应鼓励基础研发，建立适应性监管框架，为后泡沫时代的技术应用铺路；对投资 者来说，泡沫期需保持警惕，聚焦有技术护城河和清晰路径的企业，泡沫破裂后则应逆势布局核心基础 设施和优质应用企业；对企业和个人而言，当下正是学习、实验和积累AI能力的黄金时期，当技术成 本下降，提前准备的主体将率先把AI转化为生产力，构建竞争优势。 互联网泡沫的洗礼，造就了更强大务实的数字时代。如今我们站在AI的相似关口，不必畏惧泡沫存 在，而应聚焦泡沫后的未来——这是从上一场技术革命中汲取的最宝贵经验。 韩和元 1995年至2001年间，全球资本市场曾上演过这样一幕：过度投机，估值远超实际价值。最终泡沫破裂导 致市场暴跌、大量企业倒闭。金融史将其称为"互联网泡沫"。如今，AI(人工智能)领域似乎正上演相似 剧情：估值高企、资本扎堆、概念炒作泛滥 ...

Core Insights - The current narrative surrounding AI is pushing the Federal Reserve into a dilemma, where lowering interest rates could lead to dangerous outcomes, while not lowering rates could push the market into a crisis [1][4] Group 1: Federal Reserve's Dilemma - The report from TS Lombard highlights that AI could either lead to a deflationary productivity boom similar to the 1990s or push up the equilibrium interest rate (r*), resulting in contrasting monetary policy paths [1][4] - Lowering interest rates based on the expectation of AI enhancing productivity is deemed a "dead end," as the current inflation environment is less favorable than in the 1990s [1][3] - Not lowering rates could lead to a scenario where inflation resurfaces in 2026, forcing the Federal Reserve to adopt tightening policies, which could inadvertently burst market bubbles [1][4] Group 2: Historical Context and Lessons - The report discusses Alan Greenspan's strategy of "cleaning up rather than intervening," suggesting that future Federal Reserve chairs may follow this approach, especially those appointed by pro-technology leaders [2][12] - Greenspan's legacy is complex, as he initially delayed rate hikes in the 1990s due to underestimating productivity growth, but later raised rates to prevent excessive monetary policy [3][4] - The historical context indicates that simply advocating for rate cuts based on past experiences with Greenspan overlooks the nuanced challenges faced by the Federal Reserve during technological revolutions [4][12] Group 3: Key Questions Influencing Policy - Three critical questions will shape the Federal Reserve's policy path: 1. Whether large-scale capital expenditures in the tech sector are inflationary [6] 2. If AI can deliver productivity gains similar to those seen in the 1990s [10] 3. Who benefits from productivity improvements, with historical trends suggesting workers may gain more than corporations [11] Group 4: Economic Implications of AI - AI could act as a powerful deflationary force if productivity increases while wage growth remains stable, leading to lower unit labor costs and potentially lower prices for consumers [7] - Conversely, the surge in capital expenditures driven by AI may elevate the equilibrium interest rate, as higher expected returns on capital encourage significant investments [7][10] - The report notes that the potential for AI to replicate the productivity growth of the 1990s is uncertain, with estimates of AI's contribution to productivity varying widely among experts [10][11] Group 5: Future Considerations - The Federal Reserve's traditional approach of not actively bursting asset bubbles may lead to unintended consequences if inflation becomes a primary concern again [12] - The current inflation dynamics are less favorable than those in the 1990s, which could complicate attempts to replicate Greenspan's policies without risking a tech bubble [12]

Core Viewpoint - Michael Burry expresses concerns about the current AI hype, comparing it to the late 1990s internet bubble, particularly highlighting Nvidia as a potential indicator of an impending industry bubble burst [2][3]. Group 1: AI Hype and Nvidia - Burry labels the current AI trend as a "brilliant absurdity," identifying Nvidia as a central player in this bubble, akin to Cisco during the internet bubble [2]. - He draws parallels between the tech giants of the past, such as Microsoft and Cisco, and today's AI leaders, which he refers to as the "five knights": Microsoft, Google, Meta, Amazon, and Oracle [2]. - Nvidia's market capitalization has surged to approximately $5 trillion, making it the highest-valued company globally, raising concerns about its inflated valuation [5]. Group 2: Historical Comparisons - Cisco's stock price skyrocketed by 3,800% from 1995 to 2000, reaching a market cap of about $560 billion before crashing over 80% at the turn of the century, which Burry believes is a historical pattern repeating itself with Nvidia [3]. - Burry's hedge fund, Scion Asset Management, purchased over $1 billion in put options against Nvidia and Palantir, indicating skepticism about their future performance [4]. Group 3: Industry Dynamics - Concerns are raised about the interconnected financing among AI companies, with Nvidia committing significant investments to firms like OpenAI and Anthropic, creating a cycle of funding that may inflate valuations further [5]. - Nvidia's CFO refutes Burry's claims regarding the lifespan of its chips, asserting that their hardware is durable and efficient due to the CUDA software system [5]. - CEO Jensen Huang counters bubble concerns by stating that the company has not yet allocated any actual funds and that planned investments represent a small fraction of its revenue, emphasizing a long-term growth cycle in computing technology [6].

26 星期三 2025年11月 信息来源：网络 编辑-楓華 感恩节前的市场，像一幅流动的画卷，色彩斑斓又充满变数。美联储的鸽派信号、人工智能的热潮，以 及加密货币的无厘头波动，共同编织了一场复杂的金融交响。全球投资市场里的新老韭菜们既兴奋又忐 忑，仿佛置身于一个不断切换光影的舞台。 市场苦等的那只鸽子 当纽约联储主席和理事先后释放鸽派信号时，市场的情绪瞬间发生转向。从谨慎观望到试探入场，投资 者的心态如同秋风掠过水面，轻轻搅动起涟漪。纳斯达克和标普反弹，风险资产短暂获得喘息，但背后 的故事，比数字更耐人寻味： 在经济和政策的交汇点，每一次表态都像是一面风向标，让人重新思考风险与机会的边界。 大空头再发警世恒言 人工智能板块这两天的反弹引人注目，但市场的热情里，也潜伏着焦虑。迈克尔·伯里，这位准确预测 2008年金融危机的大空头，再次提醒市场：历史总喜欢重演。二十年前的互联网泡沫与今天的AI狂 潮，有着惊人的相似——投资者往往被成长的想象力所吸引，而忽略了盈利的现实。 伯里的提醒不仅是警告，也像是一面镜子，映照出投资者在泡沫与现实之间徘徊的心理。正如政策制定 者公开保证市场安全，但经验告诉我们，过度自信往往孕育 ...

Core Insights - Concerns about an artificial intelligence (AI) bubble have led to market volatility, but the risk of a systemic collapse similar to the 2000 internet bubble is considered low due to fundamental differences between AI and the internet bubble [1][8][9] - The AI industry is experiencing localized overheating, but key indicators such as CAPEX growth, debt financing ratios, and profitability need to be monitored to assess potential risks [1][8] - AI applications in various industries are expanding, with 65% of surveyed companies in mainland China planning to increase AI usage in the next 12 months, a 17 percentage point increase from the previous survey [1][2] Industry Trends - AI is expected to accelerate vertical development across different industries, integrating closely with industry characteristics and business models [2][25] - The trend of "human-machine collaboration" is becoming more pronounced, with companies reducing entry-level accounting positions while increasing the hiring of AI-skilled professionals [3][4] - The employment market is shifting from "job replacement" to "value enhancement," focusing on high-value functions that AI cannot easily replace [4][10] Challenges in AI Adoption - Companies face three main challenges in AI implementation: cost and return on investment uncertainty, technology and organizational fit, and compliance and risk management pressures [4][5][6] - 40% of surveyed companies cite financial costs and low ROI as primary challenges, with 49% of small and medium-sized enterprises (SMEs) particularly sensitive to these issues [6][8] - The complexity of integrating AI with existing technology systems poses significant challenges, especially for SMEs that often lack technical talent [5][6] Strategic Recommendations - Companies should anchor their AI investments to application value, focusing on quantifiable outcomes rather than following trends blindly [10][11] - Balancing short-term costs with long-term capabilities is crucial, with SMEs encouraged to adopt lightweight third-party AI tools initially [11][17] - Organizations should establish a governance framework for AI that encompasses data collection, model training, and application deployment to mitigate risks and ensure compliance [12][13][19] Future Outlook - AI is expected to continue its integration into various sectors, with significant applications already seen in finance, accounting, and auditing [12][13] - The "14th Five-Year Plan" in China is anticipated to drive new productivity through AI, fostering innovative digital business models [2][25] - By 2026, companies are advised to focus on AI, data analytics, and business intelligence software as key areas for technological investment [25]

近期，投资者对AI泡沫的担心持续上升。就此，在11月26日举办的中国银河证券2026年度投资策略报 告会上，中国银河证券首席经济学家、研究院院长章俊表示，AI泡沫确实存在，但会否演变成危机， 目前来看还为时尚早。章俊提到，IMF将当下与互联网泡沫进行了对比，从市盈率和投资热度等指标来 看，目前整体依然相对可控。美联储加息是刺破互联网泡沫的重要原因，但当下是美联储和全球降息周 期，政策风险可控。不过与此同时，他强调，目前全球经济的韧性远低于当年，因此指标的可比性也还 是存在不确定性。 ...

近几周，伯里加强对英伟达的审视，包括关注美国AI企业之间"循环投资"、收入确认方式以及技 术巨头如何对计算设备计提折旧。例如，据9月与美国开放人工智能研究中心（OpenAI）达成的 一项协议，英伟达计划向OpenAI投资1亿美元，后者则将使用数百万枚英伟达芯片来建设数据中 心。 美国彭博新闻社24日提出的另一个问题是，英伟达的图形处理器（GPU）会多久过时。此外还有 不少人质疑，英伟达的AI基础设施支出能否持续。 （据新华社北京11月25日电） 他强调，思科是互联网泡沫破裂时的前沿企业，其股价在 1995 年至 2000 年间飙升3800%，市值 一度约5600亿美元，成为当时全球市值最高公司。然而，千禧年之交，其股价暴跌超过80%。 伯里认为，如今历史正在美国AI热潮中重演："再次有一家'思科'处于这一切的中心……它的名字 叫英伟达。" 与25年前思科创下市值纪录类似，英伟达如今成为全球市值最高公司，估值约为5万亿美元。 美国知名空头迈克尔·伯里23日在付费订阅平台"订阅堆栈"发文，重申他对美国英伟达公司的看空 立场。此前，英伟达反驳了伯里的相关分析。 在这篇题为《泡沫的关键标志：供给侧的贪婪》的文章中，伯 ...

Core Viewpoint - Michael Burry, a well-known short-seller, reiterates his bearish stance on Nvidia, suggesting that the current AI hype in the U.S. is a "magnificent absurdity" and that Nvidia is a precursor to a potential bubble burst in the AI industry [1][3]. Group 1: Market Analysis - Burry compares the current AI boom to the internet bubble, highlighting that Nvidia is at the center of this phenomenon, similar to Cisco during the late 1990s [3]. - Nvidia has become the highest-valued company globally, with an estimated valuation of approximately $5 trillion, mirroring Cisco's market dominance in its time [3]. Group 2: Financial Performance and Concerns - Burry emphasizes the importance of examining Nvidia's revenue recognition methods and the "circular investment" among AI companies, which could indicate underlying financial instability [3]. - Questions have been raised regarding the longevity of Nvidia's GPU technology and whether its AI infrastructure spending can be sustained [4]. Group 3: Historical Context - Burry references Cisco's stock price surge of 3800% from 1995 to 2000, followed by an over 80% decline after the bubble burst, suggesting a similar fate could await Nvidia [3].