高盛最新推出了一个包含多家非科技类公司的新组合GSXUPROD，成分股涵盖金融、零售、物流、医疗和餐饮等行业，这些公司已明确将AI整合进工作 流程以降低成本、提高利润率。高盛认为，GSXUPROD组合通过AI采用和劳动生产力提升带来的基准每股收益变化潜力，高于罗素1000指数和标准普 尔500指数。 该组合今年迄今表现落后于大盘，即使剔除科技七巨头（Mag7）后仍然如此。不过，高盛股权投资组合策略团队的框架分析显示，GSXUPROD组合 通过AI采用和劳动生产力提升带来的基准每股收益变化潜力，高于罗素1000指数和标准普尔500指数。 以下是AI受益者的完整名单： 银行与保险：自动化与成本控制并进 金融机构正在大规模部署AI以提升运营效率，主要应用领域涵盖编码、机构服务、欺诈检测、信贷、催收、营销和承保。这些公司通过AI平台和虚拟助 手自动化日常任务并增强客户互动。许多金融机构将AI视为增强人力而非替代的工具，在降低成本的同时支持员工工作，部分公司正试点智能体AI以实 现更广泛整合并控制员工人数增长。 距离ChatGPT发布已过去三年，企业AI应用正从概念走向实质，高盛认为做多AI生产力受益者组合将是2026 ...

亚马逊表示，在这些基础设施下，美国的联邦政府机构将获得更广泛的AWS人工智能服务访问权限， 包括用于训练模型的SageMaker AI、用于代理部署的Bedrock、Claude模型，以及Trainium芯片和英伟达 芯片。根据媒体与调查网站SourceMaterial查阅的文件显示，亚马逊已经在全球50多个国家拥有逾900处 数据中心设施，这一运营规模远超外界普遍认知。 中信建投证券指出，近期，算力板块波动加大。复盘2023年以来的算力板块，三季报后调整比较常见， 考虑到AI带动的算力需求依然非常旺盛，调整是机会。目前，世界正处于AI产业革命中，类比工业革 命，影响深远，不能简单对比近几年的云计算、新能源等，需要以更长期的视角、更高的视野去观察。 因此，对AI带动的算力需求以及应用非常乐观。 CPO概念25日盘中强势拉升，截至发稿，德科立20%涨停，光库科技涨超17%，长光华芯、长芯博创涨 超15%，腾景科技涨超10%续创新高，中际旭创、新易盛等涨逾7%。 消息面上，科技巨头亚马逊宣布将投资高达500亿美元，用于扩展公司面向美国政府客户的AWS人工智 能和高性能计算（HPC）能力。 ...

每经编辑｜陈柯名 据央视新闻，当地时间11月24日，美国白宫发布声明表示，总统特朗普签署了一项行政命令，启动一项旨在利用人工智能（AI）变革科学研究方式、加 速科学发现的全新国家计划"创世纪计划"。 图片来源：视觉中国 声明表示，该命令指示能源部创建一个人工智能实验平台，整合美国超级计算机和独特数据资产，以生成科学基础模型并为机器人实验室提供支持。该命 令指示总统科学与技术事务助理（APST）协调这项国家计划，并整合联邦政府各部门的数据和基础设施。能源部长、总统科学与技术事务助理以及人工 智能与加密技术特别顾问将与学术界和私营部门的创新者合作，支持并加强"创世纪计划"。 特朗普在7月公布了一项全面的人工智能政策蓝图，旨在使人工智能公司更容易在美国发展，并使美国盟友更容易获得关键的硬件和软件。该蓝图鼓励美 国能源部和其他机构与私营部门和国家实验室合作，投资于"一系列科学领域的自动化云实验室，包括工程、材料科学、化学、生物学和神经科学"。它还 指示政府扩大实验室的人工智能研究和培训。 上周，美国能源部幕僚长卡尔·科表示，这表明特朗普政府认为即将到来的人工智能竞赛与"曼哈顿计划"或太空竞赛同等重要。卡尔·科说："我 ...

以下文章来源于财联社AI daily ，作者赵昊 财联社AI daily . 财联社及科创板日报旗下产品——未来已来，AI前沿，独家、深度、专业！ 根据媒体与调查网站SourceMaterial查阅的文件显示，亚马逊已经在全球50多个国家拥有逾900处数据中心设施，这一运营规模远超外界 普遍认知。 亚马逊周一（11月24日），在官网声明中写道，这项投资计划将于2026年破土动工，通过在AWS Top Secret、AWS Secret和AWS GovCloud (US) Regions建设配备先进计算和网络技术的数据中心，新增近1.3吉瓦的人工智能和超级计算能力。 亚马逊表示，在这些基础设施下，美国的联邦政府机构将获得更广泛的AWS人工智能服务访问权限，包括用于训练模型的SageMaker AI、 用于代理部署的Bedrock、Claude模型，以及Trainium芯片和英伟达芯片。 亚马逊称，这些能力将帮助美国超过11,000个政府机构开发定制化AI解决方案、优化海量数据集处理，并提高员工生产力，以支持新一代 创新。 另外，AWS还在数百个所谓"托管"（colocation）设施中部署了服务器机架。业内通常 ...

Core Viewpoint - The AI arms race has entered its 2.0 phase, shifting focus from general-purpose GPUs to self-developed ASIC chips and interconnect technologies by CSP cloud vendors [1][2] Group 1: AI Development and Investment - Since the launch of ChatGPT 3.5 in 2023, major tech companies have significantly increased investments in large model development and computing center construction, with projected Capex for Amazon, Google, Microsoft, and Meta reaching $361 billion by 2025, a year-on-year increase of over 58% [2] - Domestic companies like ByteDance, Tencent, and Alibaba are expected to exceed 360 billion yuan in Capex [2] - NVIDIA, as a leading AI chip manufacturer, has faced supply shortages, while CSPs are increasingly investing in self-developed ASIC chips for better cost-performance ratios [2] Group 2: CSP Cloud Vendors' Self-Developed ASIC Chips - Google has been developing its TPU ASIC chips since 2015, with plans for the seventh generation and innovations in OCS architecture starting from TPUv4 [3] - AWS's Trainium chip is progressing to its third generation, with notable interconnect innovations using copper cables and plans for copper backplane connections in future clusters [3] - Meta has designed its MTIA chip and has a long history of data center architecture design, including the well-known CLOS architecture [3] - Other companies like Broadcom and Marvell are actively supporting global CSP cloud vendors in data center construction [3] Group 3: Silicon Photonic Modules - Silicon photonic modules integrate various optical components into a single chip, offering low cost, low power consumption, and high integration [4] - The application scenarios for silicon photonic modules include data center communication and telecom networks, benefiting from the demand for cost-effective solutions driven by AIGC transformations [4] - The market for silicon photonic modules is projected to reach $10.3 billion by 2029, with a CAGR of 45% over the past five years, corresponding to nearly 18 million units sold [4] Group 4: Growth in Optical Communication Market - The demand for ASIC chips is expected to increase, with global shipments of 800G optical modules projected to reach 40 million units and 1.6T optical modules exceeding 7 million units next year [5] - The penetration rate of CPO is expected to reach 50% by 2029, with the OCS market projected to exceed $1.6 billion [5] - The PCIe Switch market is anticipated to reach $5 billion, while the DCI market is expected to reach $28.4 billion [5]

OpenAI 已签署协议，将从亚马逊云服务（AWS）采购价值 380 亿美元的算力资源。这是 OpenAI 首次 与云基础设施领域的领导者达成合作，也进一步表明这家估值 5000 亿美元的人工智能初创公司不再依 赖微软。 在今年之前，OpenAI 与微软一直保持着独家云服务合作关系 —— 微软于 2019 年首次对 OpenAI 进行 投资，累计投资额达 130 亿美元。今年 1 月，微软表示将不再担任 OpenAI 的独家云服务提供商，转而 调整合作模式，仅保留对 OpenAI 新算力需求的 "优先认购权"。 上周，根据微软与 OpenAI 新谈判的商业条款，微软的 "优先合作地位" 正式到期。这使得 ChatGPT 的 母公司得以更广泛地与其他超大规模云服务提供商（ hyperscalers ）建立合作。事实上，在此之前， OpenAI 已与甲骨文和谷歌达成云服务合作，但亚马逊云服务仍是目前市场份额最高的云服务提供商。 OpenAI 首席执行官山姆?奥特曼在周一的声明中表示："扩展前沿人工智能技术需要庞大且可靠的算力 支撑。我们与 AWS 的合作将强化广泛的算力生态系统，这一生态系统将为下一个人工智能时代 ...

Core Insights - OpenAI has signed a significant $38 billion computing power agreement with Amazon's AWS, marking its first major partnership since ending its exclusive cloud collaboration with Microsoft earlier this year [1][3] - This partnership allows OpenAI to expand its computing infrastructure flexibly until 2026 and beyond, reducing its dependency on Microsoft [1][6] - The agreement is part of OpenAI's broader $1.4 trillion infrastructure investment plan, with AWS serving as a critical pillar for OpenAI's ambitious AI goals [1][2] Agreement Details - The initial phase of the agreement will utilize existing AWS data centers, with plans for Amazon to build additional infrastructure for OpenAI [2][6] - OpenAI will begin migrating some computing tasks to AWS, starting with tens of thousands of NVIDIA GPUs located in the U.S. [1][2] - OpenAI has also signed infrastructure agreements totaling approximately $1.4 trillion with other companies, including NVIDIA, Broadcom, Oracle, and Google, raising discussions about potential market bubbles in the AI sector [2][6] Market Impact - Following the announcement, Amazon's stock price rose over 6% in pre-market trading, reflecting positive market sentiment towards the deal [1][3] - OpenAI's CEO emphasized the importance of large-scale, reliable computing support for advancing AI research, indicating that the partnership with AWS is a key step in enhancing the AI technology ecosystem [5][6] Competitive Landscape - Despite diversifying its partnerships, OpenAI will continue its deep collaboration with Microsoft, reaffirming a commitment to purchase an additional $250 billion worth of Azure cloud services [5][6] - AWS remains a leader in the cloud computing space, even as Microsoft and Google experience faster growth rates in their cloud businesses, with increases of 40% and 34% respectively [6][7] Future Prospects - OpenAI's collaboration with AWS positions it for a potential IPO, as the company aims to demonstrate its ability to operate independently and mature its business model [7][8] - The partnership is expected to support both AI inference and the training of next-generation models, with plans for scaling computing resources in alignment with business growth over the next seven years [6][7]

Investment Rating - The report maintains a "Buy" rating for several companies in the AI and computing sector, including 亿道信息 (001314.SZ), 唯科科技 (301196.SZ), 能科科技 (603859.SH), 嘉和美康 (688246.SH), 合合信息 (688615.SH), and 迈信林 (688685.SH) [11]. Core Insights - Qualcomm launched AI inference chips AI200 and AI250, focusing on low power consumption and high cost-effectiveness, aiming to challenge NVIDIA's dominance in the AI chip market [4][17]. - NVIDIA's GTC conference showcased its strategic vision for AI, including advancements in next-generation chip architecture, 6G communication, quantum computing, and AI infrastructure [5][27]. - AI data annotation platform Mercor completed a $350 million Series C funding round, significantly increasing its valuation and aiming to democratize high-paying knowledge work [46][48]. Summary by Sections Computing Power Dynamics - The rental prices for computing power remain stable, with specific pricing for cloud services noted [16]. - Qualcomm's AI chips are designed for data centers, featuring a unique memory architecture and modular deployment, which could reshape the competitive landscape in data center computing [17][19]. AI Application Dynamics - Perplexity's weekly traffic increased by 5.73%, indicating growing interest in AI applications [5][26]. - NVIDIA's GTC conference highlighted its comprehensive strategy for AI, addressing challenges in the computing industry and presenting solutions through parallel computing and GPU integration [29][30]. AI Financing Trends - Mercor's recent funding round reflects a growing market for AI data annotation services, with plans to expand into healthcare and legal sectors [46][48]. - The competitive landscape for AI platforms is evolving, with Mercor positioning itself as a leader in high-value expert tasks [47]. Market Performance Review - The AI application index and AI computing index showed varying performance, with notable gains and losses among specific companies [52][58]. Investment Recommendations - The report suggests focusing on companies with successful clinical AI product validations and those expanding their computing power capabilities, such as 嘉和美康 (688246.SH) and 迈信林 (688685.SH) [60].

Core Insights - OpenAI has signed a $38 billion agreement with Amazon Web Services (AWS) for computing resources, marking its first partnership with a leading cloud infrastructure provider and signaling a reduced reliance on Microsoft [1] - The deal allows OpenAI to utilize existing AWS data centers and will expand its computing capacity over the coming years, with immediate deployment of workloads on tens of thousands of NVIDIA GPUs [1][2] - This partnership is strategically significant for AWS as it aims to benefit from the competitive landscape of AI computing [1][3] Group 1: OpenAI's Partnerships and Market Position - OpenAI has recently signed agreements with major companies including NVIDIA, Broadcom, Oracle, and Google, totaling approximately $1.4 trillion, raising concerns about potential AI bubble risks [2] - Prior to this, OpenAI had an exclusive cloud partnership with Microsoft, which invested a total of $13 billion since 2019, but has shifted to a model where it retains priority purchasing rights [2] - OpenAI's CEO, Sam Altman, emphasized the need for substantial and reliable computing power for the scaling of advanced AI technologies [2] Group 2: AWS's Strategic Importance - The agreement is crucial for AWS not only due to its size but also because of its existing ties with OpenAI's competitor, Anthropic, which has received significant investment from Amazon [3] - AWS reported over 20% year-on-year revenue growth, although its growth rate is slower compared to Microsoft and Google, which achieved 40% and 34% respectively [3] Group 3: Future Developments and IPO Considerations - The current agreement specifies the use of NVIDIA chips, with potential for future inclusion of Amazon's custom Trainium chips, which are already in use by Anthropic [4][5] - OpenAI plans to scale its computing resources through AWS over the next seven years, with the potential for further expansion beyond 2026 [5] - The partnership with AWS is seen as a step towards OpenAI's preparation for an eventual IPO, indicating a move towards greater independence and operational maturity [5][6]

Core Insights - OpenAI has entered into a $38 billion (approximately 270.73 billion RMB) computing power procurement agreement with Amazon Web Services (AWS), marking its first collaboration with a global cloud infrastructure leader and reducing its reliance on Microsoft [1][3][4] - The partnership will initially utilize hundreds of thousands of Nvidia GPUs deployed in the U.S. and plans to expand computing capacity in the coming years [3] - OpenAI has announced a total of approximately $140 billion (around 997 trillion RMB) in infrastructure investment agreements with various companies, raising concerns about a potential AI bubble and resource availability in the U.S. [3] OpenAI and AWS Collaboration - OpenAI will begin running workloads on AWS infrastructure immediately, with the first phase leveraging existing data center facilities [3] - AWS will build additional infrastructure specifically for OpenAI in the future, and the collaboration is expected to enhance the overall computing ecosystem for advanced AI technologies [4][5] - OpenAI's CEO, Sam Altman, emphasized the need for massive and reliable computing power to support the scaling of frontier AI [4] Impact on Amazon - The agreement is strategically significant for Amazon, especially as AWS has invested billions in OpenAI's competitor, Anthropic [5] - AWS reported over 20% year-on-year revenue growth, surpassing analyst expectations, although Microsoft and Google Cloud's growth rates were higher at 40% and 34%, respectively [5] - The deal specifies the use of Nvidia chips, with potential future inclusion of Amazon's own Trainium chips for cost-effective options [5][6] Future Prospects - OpenAI has committed to purchasing an additional $250 billion (approximately 1.78 trillion RMB) worth of Azure cloud services from Microsoft, indicating continued collaboration despite the new partnership with AWS [4] - The infrastructure will support both inference tasks and training of next-generation models, with flexibility for OpenAI to expand its collaboration with AWS over the next seven years [6] - The partnership is seen as a crucial step towards OpenAI's potential IPO, signaling its operational independence and mature business structure [6][7]