尽管AI投资主题基础依然稳固，但巴克莱指出数据中心资本支出放缓可能成为美股最大的系统性风险。 追风交易台消息，9月25日巴克莱股票策略团队发表研报，指出 如果数据中心资本支出在未来两年下降20%，标普500指数将面临3-4%的盈利下行压力，更严重 的是估值可能下跌10-13% 。 报告指出了可能引爆这场危机的三大潜在"巨雷"： 01 AI热潮根基稳固 但并非无懈可击 首先，报告肯定了AI投资主题的坚实基础。 即便是在一个预计每年增长30%的数十万亿美元资本支出预测下，市场对算力的需求仍然远超供应。 巴克莱股票策略团队指出如果数据中心资本支出在未来两年下降20%，标普500指数将面临3-4%的盈利下行压力，更严重的是估值可能下跌10-13%。报告识别三大风险：AI模型效 率提升可能导致计算设施过度建设；电力短缺成为数据中心建设物理限制；资本支出增长超越现金流创造能力时的融资压力。 硬·AI 作者 | 鲍奕龙 编辑 | 硬 AI 高级推理模型和AI代理的普及进一步推高了需求天花板。 数据显示 ， 标普1500指数中，有十分之一的公司在财报中提及AI带来的效率提升。 技术与效率风险 ：AI模型效率的飞速提升可能导致 ...

Group 1 - The core viewpoint of the article is that the AI investment advisory market is expected to grow significantly, with a projected increase from $61.75 billion in 2023 to nearly $471 billion by 2029, representing a growth of over 600% [2][3] - Currently, about 10% of retail investors are using chatbots for stock selection, and half of the surveyed individuals are open to trying it [4][6] - An experiment reported by Finder indicated that a portfolio selected by ChatGPT achieved a remarkable return of 55%, outperforming mainstream funds in the UK market [4][6] Group 2 - Former UBS analyst Jeremy Leung mentioned that he now uses ChatGPT to guide his investment portfolio, stating that even simple tools can replicate many of his previous workflows [6][7] - Industry experts have raised warnings about the risks associated with using general AI models, emphasizing that they may misquote data and overly rely on existing narratives [6][7] - Leung cautioned that chatbots cannot access data behind paywalls, potentially missing critical analytical information [7] Group 3 - To achieve optimal results, users must provide detailed instructions to the AI, such as specifying the context or using reliable sources like SEC filings [9] - There is a concern that if investors become too complacent after easy gains from AI, they may struggle to respond effectively during market crises or downturns [9]

高盛称，快手发布的"可灵AI 2.5 Turbo"模型在保持顶级性能的同时实现近30%成本削减，确立全球AI视频生成领域顶尖 地位。该模型在与主要竞争对手对比测试中表现出色，用户偏好率最高达69%。预计2025财年可灵AI年度收入将达1.54 亿美元，高于官方指引的1.25亿美元。 作者 | 硬 AI 编辑 | 硬 AI 技术突破与成本控制的完美结合，快手AI视频大模型再次刷新行业标杆。 9月25日，据硬AI消息，高盛在最新研报中称，快手发布的"可灵AI 2.5 Turbo"模型在保持顶级性能的同时 实现了近30%的成本削减， 将快手定位为全球AI视频生成领域的顶尖玩家。 研报称，快手于9月23日发布最新的"可灵AI 2.5 Turbo"模型，据公司介绍，该模型在 文本响应、动态效 果、风格一致性和美学质量等方面取得了实质性改进 ，并进一步 增强了视频生成过程中的可控性、稳定 性和一致性 。尤其值得注意的是， 在高质量模式下(1080p)，新模型生成5秒视频的价格仅为25点，比2.1 版本便宜近30%。 高盛分析师认为，这一模型在与主要竞争对手的对比测试中表现出色，同时价格更具竞争力，有望推动用 户增长和商业 ...

Core Viewpoint - Alibaba Cloud is expected to experience significant growth in data center capacity, with annual additions projected to exceed 3 GW from 2026 to 2032, which is comparable to the entire Chinese market's expected growth in 2025 [2][6][9]. Investment and Capacity Expansion - Morgan Stanley estimates that Alibaba Cloud will add over 3 GW of data center capacity annually from 2026 to 2032, equating to the total expected growth of the Chinese market in 2025 [2][6]. - UBS predicts that Alibaba Cloud's investment intensity and sustainability guidance exceed market expectations, estimating annual capital expenditures of 100 billion to 200 billion RMB due to the anticipated 1-2 GW annual capacity growth [4][9][10]. - The total data center capacity of Alibaba Cloud in 2022 was approximately 1.5 GW, with projections indicating a tenfold increase to 15 GW by 2032, necessitating an average annual net increase of at least 1-2 GW [9]. Market Implications - The anticipated growth in Alibaba Cloud's capacity and investment is expected to provide sustained growth momentum for data center operators and infrastructure suppliers [3][7]. - The report emphasizes that Alibaba Cloud's capital expenditures alone could support the growth of the data center industry for years to come [7]. - UBS's analysis suggests that the projected capital expenditures significantly surpass current market expectations, addressing concerns about the sustainability of AI capital expenditure cycles and the intensity of spending by large-scale vendors [10].

Core Viewpoint - Alibaba's Tmall Genie Smart Home 3.0 marks a significant shift from device connectivity to "spatial intelligence," enabling homes to possess sensing and cognitive capabilities [3][14]. Group 1: Strategic Upgrade - Tmall Genie announced a comprehensive upgrade to its "spatial intelligence" strategy, focusing on three core capabilities: spatial perception, spatial understanding, and ecological service [6]. - Spatial perception involves a distributed spatial network host acting as a "nerve center," integrating AI sensors to create a network that allows for proactive responses to environmental changes [6]. - Spatial understanding leverages Alibaba's Tongyi model to develop a HomeAgent that learns family habits, enabling anticipatory actions without explicit commands [6][10]. - Ecological service is enhanced through partnerships with top brands, allowing seamless multi-device collaboration for various scenarios [6]. Group 2: Technological Breakthroughs - The new Kunlun T20S distributed spatial network host supports WiFi 7, enabling rapid scene control and local data processing for enhanced user experience [8]. - AI spatial sensors can cover large areas (up to 64 square meters) and track multiple individuals simultaneously, facilitating dynamic lighting and environmental adjustments [10]. - The system evolves from rule-based scenes to a HomeAgent that automatically adjusts settings based on environmental changes and user habits, eliminating manual configurations [10]. Group 3: Ecosystem Expansion - By 2025, Tmall Genie has expanded its ecosystem to include leading companies across various industries, establishing over 70 flagship experience stores in key markets [11]. - The establishment of the "Genie Future Home Spatial Intelligence Designer Alliance" aims to provide comprehensive solutions from technology implementation to aesthetic design [11]. Group 4: Conceptual Shift - Tmall Genie 3.0 is positioned as an "active service partner" rather than a passive tool, transforming homes into intelligent entities that understand and respond to user needs [14].

硬·AI 作者 | 赵 颖 编辑 | 硬 AI AI需求强劲支撑上，存储芯片巨头美光交了一份亮眼的成绩单，上季营收劲增46%，而且指引碾压预期再 创新高，并预期资本支出将进一步增长。 在财报发布后的财报电话会中，美光表达了乐观的预期，2025财年资本支出将大幅增长，占收入比重将达 到35%左右，2025财年预计实现数十亿美元的HBM收入；先进制程的供应非常紧张，供应紧张不仅存在 于HBM市场，也存在于非HBM市场。 美光预计，2025财年资本支出将大幅增长，占收入比重将达到35%左右，预计新建晶圆厂和HBM资本支出投资的增长将 占增长的绝大部分；先进制程的供应非常紧张，在2022年和2023 年随着资本支出的减少，以及全行业向新技术节点的转 型，晶圆产能已从峰值水平大幅下降。供应紧张不仅存在于HBM市场，也存在于非HBM市场。 公司参与者 Satya Kumar - 投资者关系 Sanjay Mehrotra - 总裁兼首席执行官 Mark Murphy - 执行副总裁兼首席财务官 电话会议参与者 以下是电话会要点： 1、美光在2024财年的资本支出为81亿美元，根据目前的资本支出和收入预期，预计 2025财 ...

硬·AI 作者 | 鲍亦龙 编辑 | 硬 AI 英特尔寻求让苹果提供投资，作为提振业务的后备计划，双方讨论了如何更加紧密地合作。 周三据媒体报道，芯片制造商英特尔已接洽苹果公司寻求投资，作为其复苏计划的一部分。 两家公司还讨 论了加强合作的可能性，但谈判仍处于早期阶段，可能不会达成协议。 苹果曾是英特尔的长期客户，但在过去五年中转向自研处理器，其最先进的芯片目前由合作伙伴台积电生 产。因此，市场普遍认为， 即便苹果投资英特尔，也不太可能在未来的设备中重新换回英特尔的处理器。 除苹果外，媒体援引知情人士称英特尔也已接洽其他公司，探讨可能的投资与合作关系。 这一潜在交易将延续英特尔近期的融资势头。英伟达上周宣布对英特尔投资50亿美元，计划在个人电脑和 数据中心芯片领域展开合作。日本软银集团上月也宣布向英特尔投资20亿美元。 消息传出后，英特尔股价盘中一度飙升8%至31.70美元，随后涨幅小幅回落收涨6.41%，但在盘后时段股 价继续延续涨势。 英特尔股价因寻求与苹果进行初步合作谈判的消息而大涨，最终收涨6.41%，盘后时段股价延续涨势。但报道称谈判尚处 早期，前景未明。市场普遍认为，即便苹果投资，也不太可能在设备 ...

新模型、Agent（智能体）应用等AI软件产品及AI服务器等基础设施硬件或是看点。阿里巴巴CEO吴泳铭表示，大模型是 下一代操作系统。AI Cloud是下一代计算机。未来全世界或只会有5-6个超级云计算平台。 作者 | 龙 玥 编辑 | 硬 AI 9月24日，阿里巴巴2025云栖大会在杭州正式开幕。 本次大会将于24日至26日举行三天，以"云智一体·碳硅共生"为主题，设置 三大主论坛"云栖前瞻""云栖技 术""无法计算的价值"，共同探讨AI、云计算与产业应用的最新趋势。 硬·AI 新模型、Agent（智能体）应用等AI软件产品及AI服务器等基础设施硬件或是看点。上证报记者获悉，阿 里云将展示超大规模集群、分布式训练、推理加速等能力，并首次展出单柜支持144个计算节点的高密度 AI服务器和新一代高性能网络架构HPN8.0。 阿里巴巴CEO吴泳铭发表主旨演讲称， 资本开支会在3800亿基础上，增加更多的投入。通义千问的目标 是打造AI时代的安卓系统。大模型是下一代操作系统。AI Cloud是下一代计算机。未来全世界或只会有5- 6个超级云计算平台。 01 吴泳铭： ASI才是开始 大模型将是下一代的操作系统 大会 ...

首个1GW容量英伟达系统计划明年下半年上线，采用Vera Rubin平台。黄仁勋称，10GW相当于多达500万GPU，等于 英伟达今年的总出货量，是去年两倍。Altman称，接下来几个月里，大家应该对OpenAI抱有很大期望。他将英伟达和微 软都称为OpenAI的"被动"投资者和"最重要的合作伙伴"。盘初曾跌超1%的英伟达股价迅速转涨，午盘曾涨超4%，收涨近 4%创历史新高。 投资OpenAI大规模部署数据中心的消息公布后，周一盘初曾跌1.1%的英伟达股价迅速转涨并在美股午盘 之初拉升，午盘涨幅曾扩大到4%以上，刷新日高时涨4.5%，收涨逾3.9%，创收盘历史新高，盘后小幅回 落。 01 "巨大"项目10GW相当于多达500万GPU 硬·AI 作者 | 李 丹 编辑 | 硬 AI 英伟达为科技企业的人工智能（AI）基础设施建设狂潮又添了一把火。 美东时间9月22日周一，英伟达与OpenAI宣布，签署意向书确立达成战略合作，OpenAI将借此利用英伟 达的系统打造和部署至少10千兆瓦（GW）的AI数据中心，使用数百万块英伟达的图形处理器（GPU）训 练和部署OpenAI的下一代AI模型。 为支持上述里程碑式的战 ...

Core Viewpoint - The article discusses the remarkable stock performance of traditional hardware companies in the context of the AI boom, questioning whether this is a genuine reflection of the AI infrastructure investment or a sign of an impending market bubble [1][2]. Group 1: Stock Performance of Traditional Hardware Companies - Seagate Technology has surged by 156%, becoming the best-performing stock in the S&P 500, while Western Digital and Micron Technology have seen increases of 137% and 93%, respectively [2][4]. - These storage device manufacturers are benefiting from substantial investments by large tech companies in AI infrastructure, with hundreds of billions allocated annually for semiconductors, networking equipment, and data center power [4][5]. - Despite the stock price increases, the valuations of these companies remain relatively reasonable, with Western Digital's price-to-earnings (P/E) ratio below 6, and Seagate and Micron around 10, compared to the S&P 500's expected P/E of 23 [4][5]. Group 2: Market Sentiment and Analysis - There is a divide on Wall Street regarding this phenomenon; bulls see it as a spillover effect of AI computing demand, while bears warn it may indicate a stock market bubble [2][6]. - Analysts have raised target prices for Seagate, with Benchmark Co. setting a target of $250, indicating a potential upside of over 13% from its recent closing price [5]. - The rapid stock price increases have outpaced analyst expectations, with Seagate trading over 20% above the average target price, Western Digital over 10%, and Micron slightly above expectations [7][8]. Group 3: Broader Market Trends - The AI wave has also driven stock price surges in other traditional industries, such as Vistra Corp, which has risen 53%, and Broadcom, which has reached a market cap of $1.6 trillion [9]. - Oracle has become the tenth largest company in the S&P 500 due to demand for cloud computing services, experiencing a 36% single-day surge following its earnings report [9][10]. - Concerns are raised about the potential overvaluation of AI-related products, with warnings that any rapidly rising asset could become a cautionary tale [10].