神译局是36氪旗下编译团队，关注科技、商业、职场、生活等领域，重点介绍国外的新技 术、新观点、新风向。 编者按：关于AI的又一个泡沫论。原作者以GPT-5的投入产出比来论证人工智能是泡沫，并认为金融圈 和科技圈正在打造新的"风口"来出逃。文章来自编译。 投机主宰世界。但过去不是这样的。从20世纪80年代到2008年间，情况发生了变化。投资者意识到，他 们通过炒作获得的回报远比从任何合法业务获得的回报都要多得多。毕竟，这是信息时代，而信息很容 易被操纵和商品化。这导致了互联网泡沫、2008年的信贷危机、2016-2017年的加密货币泡沫、2020年 底至2021年的加密货币泡沫，以及2022年的NFT泡沫，而最近的一股热潮则是人工智能泡沫。事实上， 全球近一半的私人投资正涌入人工智能领域，而对人工智能的投机是标普500指数近期增长的主要驱动 力。但是，就像其他泡沫在灾难性破灭前一样，人工智能泡沫也显示出即将破裂的迹象。然而，金融圈 和科技圈的那帮家伙已经吸取了教训，正在打造下一个"风口"，准备卷走我们所有的钱，好在他们不可 避免地需要抽身时，能安然离场。只可惜，这个新风口跟人工智能相比更是死胡同一个。 所以，人 ...

星球中每个交易日早上都会推送前一天晚上，外资投行/券商对国内行业/公司的分析、各大媒体上跟中国相 关的半导体和AI的资讯总结，可以理解成中国版的TMT Breakout。一般在早上9点之前发到星球。 我们最近这段时间会在当天到公众号中先公开这个China TMT Breakout的内容。很多国外的报导，在第二 天国内的媒体也会相继转发，但星球中看到这些信息都时间肯定要早于大部分网友。 Global 1、 Intel公布财报，收入为 137 亿美元，同比增长 3%——超过华尔街预期的 131 亿美元。英特尔预计第四季度的收入将 在 128 亿美元到 138 亿美元之间，基本符合市场普遍预期。这也得益于特朗普政府、英伟达和日本软银的大规模投资带 来的动力。 2、 Anthropic 与谷歌云达成重磅 AI 芯片交易，获得 100 万台谷歌云芯片的使用权，用于训练和运行其人工智能模型，谷 歌已向 Anthropic 投资超过 30 亿美元，明年将通过其定制芯片——张量处理单元（TPUs）——为这家初创公司上线超过 一吉瓦的人工智能计算能力。亚马逊是这家初创公司的"主要"云服务提供商，也是其重要投资者。它已向 Ant ...

Core Viewpoint - The discussion centers around whether current AI investments are in a "bubble," with Azeem Azhar arguing that while there are concerns, the situation does not meet the strict definition of a bubble [1][5]. Economic Pressure - Data center construction significantly contributes to the US GDP, but it has not reached historical bubble levels. Azeem Azhar identifies a threshold where investment as a percentage of GDP becomes concerning, noting that around 2% is "tricky" and 3% is "troublesome" [2][22]. - The construction of data centers is seen as a positive economic driver, creating jobs in various sectors, despite some political tensions arising from local opposition to such projects [2][23]. Industry Pressure and Revenue Growth - There is a notable disparity between AI-related capital expenditures (approximately $370-400 billion) and AI-related revenues (around $60 billion), indicating a sixfold gap [2][26]. - Azeem acknowledges this gap is concerning but emphasizes that revenue typically lags behind infrastructure investment in technology sectors. He suggests that achieving continuous annual revenue growth of about 100% in the coming years is crucial [2][27]. Valuation Heat - The stock prices of leading AI companies have surged, with AI-related firms contributing significantly to the S&P 500's performance. Azeem differentiates the current situation from the internet bubble, noting that today's financing is primarily equity-based rather than debt-based [2][33][34]. Financing Quality - Azeem highlights that a significant portion of future data center capital expenditures (estimated at $3 trillion over three years) will need to be financed through private credit and other off-balance-sheet structures, raising concerns about transparency and potential risks [3][39]. - The quality of financing is critical, as historical data shows that poor financing quality has often been a precursor to market collapses [3][40]. Summary of Indicators - Azeem outlines five key indicators to assess the AI investment landscape: economic pressure, industry pressure, revenue growth, valuation heat, and financing quality. He emphasizes that the most critical indicator is revenue growth, which must keep pace with capital expenditures to avoid a bubble scenario [1][41].

公众号记得加星标⭐️，第一时间看推送不会错过。 来 源： 内 容 编译自 semiengineering 。 我们正处于人工智能泡沫之中吗？有人将其与1999年至2000年互联网泡沫的兴衰相提并论。 在互联网泡沫破灭期间，许多高科技公司的估值飙升了10倍，随后又大幅缩水。纳斯达克综合指数的 市盈率峰值曾达到200倍！还记得Webvan吗？它于1999年11月上市，估值80亿美元，19个月后就申 请破产了。它当时只是投机行为，既没有盈利，也没有增长。 如 今 ， 估 值 大 幅 上 涨 ， 营 收 、 盈 利 和 增 长 前 景 也 同 样 强 劲 。 然 而 ， 市 场 热 情 正 逐 渐 转 向 像 Coreweave 这样存在风险的公司。 纳斯达克综合指数9月底的市盈率为33倍。亚马逊、苹果、谷歌、微软、Meta和台积电的市盈率在 27-39倍之间。英伟达市盈率高达52倍，但其利润丰厚，是人工智能领域的领头羊，并且还将持续一 段时间。AMD市盈率高达97倍，这说明他们押注于未来几年能够抢占GPU市场的显著份额。截至本 文撰写时，AMD凭借其收购OpenAI的交易，市盈率已跃升至140倍。只有特斯拉的市盈率 ...

AI 帝国的雄心勃勃必然招致批评与抵抗。 OpenAI 的 CEO 山姆·奥尔特曼（Sam Altman）又一次陷入舆论漩涡，这次是因为 ChatGPT 要给成年人放开情色内容。 简单说，12 月的 ChatGPT 会采取美国电影分级制度一样的策略给内容划分边界，给成年用户提供更多自由，用对待成年人的方式对待成年人，而对于青 少年用户则不会放宽相关政策，ChatGPT 会优先考虑安全，而不是隐私与自由。 评论区数千条留言中，充斥着对他个人和 OpenAI 的质疑和批评，他们大都带着同一个话题标签#keep 4o，因为他们不喜欢现在太理性不够共情的 GPT-5 模型，希望回到曾经与之建立深厚情感连接的 GPT-4o模型。 奥尔特曼并非没有注意到这种汹涌的民意，几周后推出的新版 ChatGPT，就会让 ChatGPT 像过去的 4o 一样有个性。而且在用户的要求下，ChatGPT 将能 像真人一样对话、使用丰富的表情符号，或者扮演朋友的角色。 奥尔特曼的此番言论立刻在社交媒体上引发轩然大波。今年 8 月 GPT-5 发布，大量用户对其取消 GPT-4o 表达了不满，如今这一情绪再次爆发。 但这没有能平息这一起风波 ...

Summary of Key Points from the Conference Call Industry Overview - The focus is on the energy sector, particularly on companies with no revenue that are experiencing inflated valuations due to speculation related to artificial intelligence (AI) [5][18]. Core Insights and Arguments - There is a significant concern that the real excess in market valuations is occurring within energy stocks rather than technology stocks, which are often profitable [4][5]. - A group of non-revenue-generating energy companies has collectively reached a market capitalization exceeding $45 billion, driven by expectations that tech companies will eventually purchase power from them [5][6]. - Oklo, a nuclear startup backed by OpenAI's CEO Sam Altman, has seen its shares rise approximately eightfold this year, resulting in a market cap of around $26 billion, making it the largest U.S. public company with no revenue in the past 12 months [6][7]. - Analysts predict that Oklo will not generate substantial revenue until 2028, as it is still in the development phase of small modular nuclear reactors [7][18]. - Another zero-revenue company, Fermi, debuted with a valuation of roughly $19 billion, making it one of the largest no-revenue companies at IPO, alongside Rivian and Corvis [8][9]. - Fermi plans to build 11 gigawatts of power capacity but has only secured equipment for 5% of this goal and lacks binding customer contracts [10][12]. Additional Important Points - Companies developing smaller "micro-modular" nuclear reactors are also seeing high valuations despite not generating revenue, with Nano Nuclear Energy valued at over $2 billion and Terra Innovatum exceeding $1 billion [13]. - Some companies, like NuScale Power and Plug Power, generate revenue but are not expected to turn a profit for several years, with projections indicating profitability may not occur until 2030 [15]. - The speculative nature of investments in energy companies is partly due to the high valuations of profit-generating firms, which have seen significant stock price increases [16]. - There is a historical precedent for zero or minimal revenue companies failing to deliver on their promises, as seen with electric vehicle startups that went public in 2020 [17]. - If the AI bubble bursts, energy companies without revenue are likely to experience the most significant declines, as they lack financial buffers [18].

Core Viewpoint - The article argues that the current state of artificial intelligence (AI) is not a bubble, but there are potential danger signals that need to be monitored through a framework of five indicators [1][2][6]. Group 1: Definition and Historical Context of Bubbles - Bubbles are not just financial phenomena but also cultural products, often associated with greed and folly [7]. - Historical examples of bubbles include the South Sea Bubble, the Roaring Twenties stock market, and the 2008 housing market crash, each characterized by overvaluation and subsequent collapse [9][10]. - The article defines a bubble as a situation where stock values drop by 50% from their peak and remain low for at least five years [10][13]. Group 2: Current AI Investment Landscape - Since the launch of ChatGPT, capital expenditures by large-scale cloud service providers have more than doubled, raising questions about the sustainability of such spending [14][16]. - Morgan Stanley predicts that AI infrastructure spending will reach $3 trillion by 2029, indicating significant investment momentum [17]. Group 3: Five Indicators Framework - The five indicators to assess the AI landscape are: 1. Economic Pressure: Evaluates whether current investment levels are distorting the economy [18]. 2. Industry Pressure: Assesses if industry revenues align with capital expenditures [30]. 3. Revenue Growth: Measures the speed of revenue growth relative to investment [35]. 4. Valuation Heat: Analyzes how high valuations are compared to historical standards [39]. 5. Quality of Capital: Examines the source and structure of funding supporting the industry [46]. Group 4: Economic Pressure - Current AI investment is at approximately 0.9% of U.S. GDP, projected to rise to 1.6% by 2030, indicating it is currently in the green zone but may soon enter the yellow zone [23][27]. Group 5: Industry Pressure - The capital expenditure to revenue ratio for generative AI is currently six times, indicating significant pressure, but this is not yet a warning sign as demand for AI services remains high [33]. Group 6: Revenue Growth - Generative AI revenue is expected to grow significantly, with estimates suggesting it could reach $1 trillion by 2028, indicating strong growth potential [38]. Group 7: Valuation Heat - Current market valuations are not as extreme as during the internet bubble, with the Nasdaq's P/E ratio around 32, which is lower than the peak of 72 during the internet boom [42][44]. Group 8: Quality of Capital - The quality of capital in the AI sector appears stable, with major companies generating substantial cash flow to support investments, although there are concerns about future funding gaps [49][51]. Group 9: Conclusion - The analysis suggests that generative AI is in a demand-driven, capital-intensive growth phase rather than a bubble, but vigilance is required as certain indicators may signal a shift towards instability in the future [52][54].

Core Insights - Oracle's revenue for Q1 FY2026 increased by 12% to $14.9 billion, with cloud computing revenue growing by 28% to $7.2 billion, while software revenue declined by 1% to $5.7 billion, indicating mixed performance [1] - Oracle signed contracts worth billions with three clients in the first quarter, and expects to secure more multi-billion dollar contracts in the coming months, with remaining performance obligations (RPO) potentially exceeding $500 billion [1] - Following the earnings report, Oracle's stock surged by nearly 36%, marking its largest single-day increase ever, adding $244 billion to its market capitalization, which reached $922 billion [1] Company Developments - The market is increasingly betting on companies investing heavily in AI and building data centers, with Nvidia and OpenAI leading the charge alongside Oracle [2] - Nvidia is investing approximately $50 billion in Intel and plans to add around $100 billion to its collaboration with OpenAI, reflecting a strong upward trend in AI semiconductor and infrastructure stocks [2] Market Sentiment - There is growing concern among investors and industry professionals about a potential AI bubble, which could pose a significant risk to the global economy [5] - OpenAI's CEO, Sam Altman, acknowledged the presence of some bubble-like conditions in the AI sector but differentiated OpenAI's genuine technological advancements and business progress from the broader market trends [7][8] Industry Dynamics - OpenAI is at the center of complex collaborations with major tech companies, including Nvidia and AMD, which are intertwined through various investment and procurement agreements [9] - The rapid rise in valuations of AI tech companies is partly attributed to "financial engineering," raising concerns about the sustainability of these valuations [10] Supply Chain Insights - TSMC reported better-than-expected earnings and raised its revenue growth forecast for 2025 to nearly 35%, indicating strong demand for AI-related products [11] - TSMC's chairman noted robust demand signals from AI clients, reinforcing confidence in the long-term growth of AI technologies [11] Historical Context - The current surge in AI investment is reminiscent of the late 1990s internet bubble, but experts suggest it may not lead to systemic risks [13] - The historical context of the internet bubble highlights the potential for over-investment outpacing actual demand, which could lead to a similar scenario in the AI sector [14]

Group 1: Oracle's Financial Performance - Oracle's revenue for Q1 FY2026 increased by 12% to $14.9 billion, with cloud computing revenue growing by 28% to $7.2 billion [1] - The company signed contracts worth billions with three clients in the first quarter, and the remaining performance obligation (RPO) may exceed $500 billion [1] - Following the earnings report, Oracle's stock surged by nearly 36%, marking the largest single-day increase in its history, adding $244 billion to its market capitalization [1] Group 2: AI Market Dynamics - Concerns are rising among investors and entrepreneurs about a potential AI bubble, which could become a global economic risk [2] - OpenAI's CEO Sam Altman acknowledged some areas of AI may be experiencing bubble-like conditions, while asserting that OpenAI itself is making genuine progress [4][5] - The intertwining relationships among major tech companies, including OpenAI, NVIDIA, and AMD, are creating complex financial arrangements that may distort true demand [5][6] Group 3: Semiconductor Industry Insights - TSMC reported better-than-expected earnings and raised its revenue growth forecast for 2025 to nearly 35%, indicating strong AI demand [6][7] - TSMC's chairman noted robust demand from AI clients, with a significant increase in processing requirements for large language models [7] Group 4: Historical Context and Future Outlook - The current surge in AI investment is reminiscent of the late 1990s internet bubble, but experts suggest it may not pose a systemic risk [8] - The infrastructure built during the internet bubble laid the groundwork for future technological advancements, similar to the current AI landscape [9] - Companies face a dilemma between expanding production capabilities and managing costs, with potential risks associated with overestimating AI demand [9]

Core Viewpoint - The article discusses the existence of multiple bubbles in the AI sector, highlighting the potential risks and opportunities for companies involved in AI investments and implementations [3][4][5]. Group 1: Types of Bubbles - The first bubble identified is an asset or speculative bubble, where AI-related companies like Nvidia and Tesla have inflated valuations, with Nvidia's P/E ratio at 50 and Tesla's at 200 despite revenue declines [3][4]. - The second bubble is an infrastructure bubble, characterized by massive investments in AI infrastructure without guaranteed future demand, reminiscent of historical overbuilding in the railroad and internet sectors [4]. - The third bubble is a hype bubble, where the promises of AI technology exceed its current capabilities, with a study indicating that 95% of AI pilot projects fail to deliver returns [4][7]. Group 2: Implications for Companies - Companies are advised not to panic in response to the bubble discussions, as the speculative and infrastructure bubbles may not directly impact most organizations [6]. - The hype bubble, however, presents a critical insight: the failure of many AI projects is often due to incorrect application rather than a lack of value in AI itself [7][8]. - Historical parallels are drawn to the internet bubble, where despite the collapse, companies that focused on building value through technology thrived [8]. Group 3: Value Creation Strategies - Successful companies should adopt a problem-oriented approach to identify friction points within their operations that AI can address [9]. - A balanced portfolio of AI initiatives should be developed, considering short-term and long-term investments, with a focus on integrating AI solutions across business functions [9][10]. - The key to thriving in the AI landscape is a systematic approach to value extraction, emphasizing clear objectives and immediate action [10]. Group 4: Opportunities Amidst the Bubble - The AI bubble may present unique opportunities for pragmatic practitioners, such as access to abundant venture capital and talent, as well as lower costs due to overcapacity in infrastructure [11]. - Companies can strategically leverage the bubble to acquire tools and technologies at reduced prices, while others bear the capital risks [11][12]. - The distraction caused by bubble discussions can provide a competitive advantage for companies that continue to focus on systematic AI implementation [12].