盘前要闻 情绪影响：正面 英伟达罕见在股价大跌后主动发声，公开宣称其图形处理器(GPU)领先行业整整一代，直接回应市场对 其在人工智能(AI)芯片领域主导地位可能受到谷歌挑战的担忧。隔夜英伟达股价一度重挫逾7%，最终 收跌约2.6%，创两个多月收盘新低。 3、特朗普称俄乌和平协议已非常接近达成，美官员称仍有一些条款需要讨论 1、国家航天局印发行动计划，目标2027年基本实现商业航天高质量发展 类型：行业 情绪影响：正面 国家航天局印发商业航天高质量安全发展行动计划(2025—2027年)，其中提出，到2027年，商业航天产 业生态高效协同，科研生产安全有序，产业规模显著壮大，创新创造活力显著增强，资源能力实现统筹 建设和高效利用，行业治理能力显著提升，基本实现商业航天高质量发展。 类型：宏观 2、英伟达大跌后罕见安抚市场：我司GPU比谷歌AI芯片领先一代 类型：公司 4、阿里Q3营收2478亿超预期，公司称AI没有泡沫、3800亿之外会加大投入 类型：公司 情绪影响：中性 阿里巴巴隔夜美股跌超2%，公司Q3营收2478亿，同比增长4.8%，略高于市场预期，但调整后净利润同 比大降72%至103.5亿，盈利能力 ...

转自：新华社 伯里提到，在互联网泡沫时期，美国技术行业的特点是"高利润大盘股，包括那个时代所谓的'四大巨头'——微软、英特尔、戴尔和思科"。 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。 新华财经北京11月25日电 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，伯里将当前美国人工智能（AI)热潮称为"一场壮丽的荒唐之举"，并点名英伟达是他预期美国AI 行业泡沫破裂的前兆。 他写道："荒唐也能赚钱。创造性的破坏和疯狂的荒唐正是美国成为世界创新中心的原因。公司可以不断创新至死，而更多公司会不断涌现，重复同样的事 情。有时，新公司只是原公司转型后的产物。 ...

美国知名空头"死咬"英伟达 在这篇题为《泡沫的关键标志：供给侧的贪婪》的文章中，伯里将当前美国人工智能(AI)热潮称为"一 场壮丽的荒唐之举"，并点名英伟达是他预期美国AI行业泡沫破裂的前兆。 他写道："荒唐也能赚钱。创造性的破坏和疯狂的荒唐正是美国成为世界创新中心的原因。公司可以不 断创新至死，而更多公司会不断涌现，重复同样的事情。有时，新公司只是原公司转型后的产物。" 伯里提到，在互联网泡沫时期，美国技术行业的特点是"高利润大盘股，包括那个时代所谓的'四大巨 头'——微软、英特尔、戴尔和思科"。 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。 卜晓明 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。 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 ...

智通财经APP获悉，Counterpoint Research的报告称，英伟达(NVDA.US)计划在其AI服务器中使用智能 手机式内存芯片，这可能导致服务器内存价格在2026年底前翻倍。 报告补充道，过去两个月，全球电子供应链受到传统内存芯片短缺的影响，原因是制造商将重心转向了 适配AI应用半导体的高端内存芯片。 英伟达定于11月19日收盘后发布第三季度财报。 Counterpoint指出，一个新问题正在酝酿。英伟达近期决定更换所用内存芯片类型，以降低AI服务器的 功耗成本，将服务器常用的DDR5(第五代双倍数据速率内存)换成LPDDR(低功耗双倍数据速率内存)。 报告提到，由于每台AI服务器所需的内存芯片数量远超智能手机，这一变更预计将催生行业难以应对 的突发性需求。 "未来更大的风险在于高端内存领域。英伟达近期转向LPDDR，意味着它将成为规模堪比大型智能手机 厂商的客户——这对供应链来说是一次重大转变，而供应链难以轻易消化如此规模的需 求。"Counterpoint表示。 这家研究公司预计，到2026年底，服务器内存芯片价格将翻倍。 服务器内存价格上涨可能会增加云服务提供商和AI开发商的成本。目前数据 ...

Core Viewpoint - AMD has outlined its growth outlook for the next 3-5 years, with expectations of achieving earnings per share (EPS) of $20 or more, driven by significant opportunities in the data center market [1] Group 1: Market Opportunities - The data center market is projected to have a compound annual growth rate (CAGR) of 40%, reaching a size of $1 trillion by 2030 [1] - This market is part of the total addressable market (TAM) for silicon, which includes CPUs, GPUs, high bandwidth memory (HBM), and networking silicon products [1] Group 2: Financial Projections - AMD is expected to achieve a revenue CAGR of approximately 35% [1] - Gross margins are anticipated to improve to 55%-58% as AI GPUs are optimized for mass production [1] - The operating leverage effect is projected to push earnings before interest and taxes (EBIT) margins above 35% and free cash flow (FCF) margins to 25% [1] Group 3: Analyst Insights - BNP Paribas analyst David O'Connor has set a target price of $300 for AMD, citing advancements in server CPU and AI accelerator markets as key drivers for future earnings [1] - If AMD achieves an EPS of around $20 in the next 3-5 years, a price-to-earnings (P/E) ratio of 25 could lead to a reasonable valuation of $500 in the coming years [1] - AMD's stock price increased nearly 9% on the day of the announcement, with a year-to-date gain exceeding 110% [1]

Market Overview - US stock index futures are all up ahead of the market opening, with Dow futures rising by 0.24%, S&P 500 futures up by 0.87%, and Nasdaq futures increasing by 1.43% [1] - European indices are down, with Germany's DAX down by 0.47%, UK's FTSE 100 down by 0.32%, France's CAC40 down by 0.26%, and the Euro Stoxx 50 down by 0.34% [2][3] - WTI crude oil prices fell by 0.38% to $60.34 per barrel, while Brent crude oil dropped by 0.40% to $64.11 per barrel [4] Economic Events - The US stock market will enter winter time next week, adjusting trading hours, with pre-market trading from 5:00 PM to 10:30 PM Beijing time and regular trading from 10:30 PM to 5:00 AM Beijing time [4] Trade Relations - US Treasury Secretary stated that the US-China trade agreement could be signed as early as next week, with both sides agreeing to enhance economic cooperation [5] Corporate Earnings - ExxonMobil reported Q3 adjusted profits of $8.1 billion, exceeding analyst expectations, driven by increased production in Guyana and the Permian Basin [7] - Chevron's Q3 net profit was approximately $3.6 billion, a 20% year-over-year decline, but adjusted earnings per share surpassed expectations at $1.85 [8] - Apple reported Q4 earnings of $102.5 billion, a 7.9% year-over-year increase, with earnings per share of $1.85, exceeding analyst forecasts [9] - Western Digital's Q3 revenue grew by 27.4% to $2.82 billion, surpassing expectations, with adjusted earnings per share of $1.78 [10] - Amazon's Q3 net sales increased by 13% to $180.2 billion, with AWS sales growing by 20%, marking the fastest growth in three years [11] - Vale's Q3 revenue rose by 9% to $10.42 billion, with net profit increasing by 78% to $2.744 billion, exceeding analyst expectations [12] Strategic Partnerships - NVIDIA announced collaborations with Samsung, SK Group, and Hyundai to build AI factories in South Korea, supplying over 260,000 accelerator chips [13] - Nokia's stock surged after NVIDIA agreed to invest $1 billion, positioning Nokia as a key player in the European AI sector [14]

Core Viewpoint - Nvidia and Deutsche Telekom are preparing to announce a plan to build a €1 billion data center in Germany, aimed at enhancing AI infrastructure in Europe [1][2] Group 1: Project Details - The project will involve a joint investment of approximately $1.2 billion (€1 billion) from Nvidia and Deutsche Telekom, with SAP as a key customer [1] - The data center is expected to utilize 10,000 advanced GPUs, which is significantly smaller in scale compared to other global data center projects [2] - The announcement is anticipated to be made next month in Berlin, involving executives from Nvidia, Deutsche Telekom, SAP, and the German Minister for Digital Affairs [1][2] Group 2: Industry Context - European policymakers and tech executives have been discussing the need for Europe to develop its own AI ecosystem to compete with the US and China [1] - The EU has previously announced a €200 billion plan to support AI development, aiming to triple the region's AI computing power within 5 to 7 years [3] - Concerns have been raised regarding the slow progress of local AI infrastructure and the need for quicker actions to support businesses while ensuring data remains within Europe [2]

Core Insights - OpenAI's valuation has reached $500 billion, with significant investments and partnerships in the AI infrastructure space, including agreements worth hundreds of billions of dollars [1][2] - Nvidia's commitment to invest up to $100 billion in OpenAI has contributed to record highs in the Nasdaq and S&P 500 indices [1] - OpenAI has secured a $300 billion computing power procurement agreement with Oracle as part of a $500 billion infrastructure project funded by SoftBank [1] - CoreWeave has agreed to provide up to $22.4 billion in AI computing power to OpenAI, increasing from an initial $11.9 billion announced earlier this year [1] - Analysts express concerns over OpenAI's reliance on a complex network of infrastructure partners and the sustainability of its investment model [2][5] Investment Dynamics - Nvidia's investment not only provides equity in OpenAI but also revenue opportunities, highlighting the intertwined financial relationships [3] - OpenAI's CFO indicated that the company expects to generate $13 billion in revenue this year, emphasizing the need for bold investments in infrastructure to support AI technology growth [4] - Some analysts warn that the financing model between OpenAI and Nvidia resembles the vendor financing seen during the early 2000s internet bubble, raising concerns about potential market instability [5][6] Future Projections - Bain & Company predicts that global computing power demand could reach 200 gigawatts by 2030, necessitating annual investments of approximately $500 billion to build sufficient data centers [6] - To cover these costs, AI companies collectively need to generate $2 trillion in annual revenue, indicating a significant funding gap of $800 billion even with full investment in cloud services and data centers [6] - OpenAI's leadership remains optimistic about the necessity of large-scale infrastructure investments, asserting that current efforts are just the beginning of what is required for AI proliferation [6]

在AI热潮的大部分时期，这类担忧始终不同程度地围绕着英伟达。根据市场数据平台PitchBook的数 据，2024年英伟达参与了逾50笔AI企业风险投资交易，且今年投资笔数有望突破这一数字。而这些获 投企业(包括AI模型开发商、云服务商等)，随后会用获得的资金采购英伟达价格高昂的图形处理器 (GPU)。 但拉斯贡指出，对OpenAI的投资规模"似乎令其他所有投资都相形见绌"。他在报告中称，这笔交易"可 能会让此类担忧比以往任何时候都更强烈，且(或许有理由)让人质疑这一举措背后的根本逻辑"。拉斯 贡补充道，英伟达已表示，对OpenAI的投资不会被用于"直接采购"英伟达产品。 微软(MSFT.US)、亚马逊(AMZN.US)等其他科技巨头，也会对头部AI初创公司进行战略投资，以推动自 身云计算业务发展。但英伟达在AI生态中占据独特地位：其主导着先进芯片市场，而这类芯片是训练 尖端AI模型的核心设备。因此，英伟达堪称迄今为止AI热潮中最大的单一受益者。 英伟达与OpenAI的合作，还发生在行业充满不确定性的节点。如今，行业内外越来越多人意识到，AI 领域可能面临类似25年前互联网泡沫破裂的风险。OpenAI首席执行官 ...

英伟达CEO黄仁勋22日接受美国消费者与商业频道(CNBC)采访时表示，10吉瓦相当于400万至500万块 图形处理器(GPU)，约等于英伟达今年的出货总量，是去年的两倍。 中新网9月23日电(记者 张乃月)据路透社报道，当地时间22日，美国芯片巨头英伟达宣布将向人工智能 企业OpenAI投资最高1000亿美元，双方将共建大规模数据中心。 报道援引接近OpenAI的知情人士称，OpenAI计划以现金方式向英伟达购买芯片，后者则会通过投资， 获得OpenAI的非控股股份。 根据双方签署的意向书，OpenAI将利用英伟达系统建设并部署至少10吉瓦的人工智能数据中心，合作 细节会在未来数周内敲定。这一耗电量，相当于800多万美国家庭的用电需求。 路透社认为，此举凸显了各大科技巨头在开发先进人工智能系统方面的利益日益重叠。这笔投资也为 OpenAI提供了购买先进芯片所需的资金和渠道，这些芯片对OpenAI在日趋激烈的竞争环境中保持实力 至关重要。 消息公布后，英伟达股价当天收盘上涨近4%。 OpenAI首席执行官奥尔特曼在声明中表示，"一切始于算力。计算基础设施会成为未来经济的基础，我 们将利用与英伟达共建的体系，既 ...