Core Insights - Microsoft, Nvidia, and Anthropic have formed a strategic partnership, creating a tightly-knit "AI alliance" with mutual investments and resource sharing [1] - Microsoft will invest up to $5 billion in Anthropic, while Nvidia will invest up to $10 billion, leading to Anthropic's valuation soaring to $350 billion [1] - Anthropic will purchase $30 billion worth of Azure computing power from Microsoft, with additional commitments for up to 1 gigawatt of computing capacity based on Nvidia's AI systems [1][2] Group 1: Partnership Details - This marks the first deep technical collaboration between Nvidia and Anthropic, focusing on optimizing Anthropic's products for performance and efficiency [2] - Anthropic's initial commitment includes acquiring up to 1 gigawatt of computing power, utilizing Nvidia's advanced architectures [2] - The partnership allows Anthropic to access Nvidia's latest AI systems, including the Grace Blackwell and Vera Rubin architectures [2] Group 2: Microsoft and Anthropic Collaboration - Microsoft is expanding its collaboration with Anthropic to provide broader access to the Claude model for enterprise users [3] - Azure AI Foundry customers will have access to Anthropic's Claude models, enhancing Microsoft's AI offerings [3][4] - Microsoft will continue to integrate Claude into its Copilot product suite, maintaining its core partnership with OpenAI while diversifying its AI model offerings [4] Group 3: Anthropic's Growth and Infrastructure - Anthropic, founded by former OpenAI employees, is recognized in sectors like finance and healthcare, with a recent valuation of $1.83 trillion after raising $13 billion [6] - The company plans to invest $50 billion in building custom data centers across the U.S. to support AI development [6] - Anthropic has also secured a deal with Google for up to 1 million AI chips, significantly enhancing its computing capabilities [6] Group 4: Market Concerns - The announcement of this partnership comes amid rising skepticism about the AI investment boom, with Nvidia and Microsoft's stock prices dropping nearly 3% on the announcement day [7] - Concerns about a potential "AI bubble" are prevalent, with 45% of fund managers viewing it as a significant market risk [8] - The cyclical investment model among tech giants raises questions about the sustainability and profitability of AI products [8]

美东时间18日周二临近美股午盘，谷歌正式发布备受期待的该司迄今最强大人工智能（AI）模型 Gemini 3，并于发布首日立即在谷歌搜索、Gemini应用程序App及多个开发者平台同步上线，在多个盈 利产品中投入使用。 在衡量人工智能模型性能的几个热门行业排行榜上，Gemini 3均领先地位。其中在全球AI模型LMArena 排行榜上以1501分的历史最高分登顶，在衡量通用推理能力的Humanity's Last Exam基准测试中获得 37.5%的最高分，超越了此前由GPT-5 Pro保持的31.64%纪录。该模型还在数学、多模态理解和事实准确 性等多个维度刷新了行业标准。 此外，被谷歌称为最佳vibe coding和智能体编码模型，推理能力有"巨大飞跃"，是"世界上最好的多模 态理解模型"、谷歌最安全模型。除了Gemini 3 Pro的预览版，谷歌还发布增强推理模式Deep Think。 思泉新材：公司深耕散热行业多年，终端品牌有谷歌、ABB、伟创力、比亚迪、富士康等。 *免责声明：文章内容仅供参考，不构成投资建议 *风险提示：股市有风险，入市需谨慎 分析认为，主要厂商在视频生成、推理能力、架构创新及服务普 ...

点击下方 卡片 ，关注" 具身智能 之心 "公众号 作者丨 Physical Intelligence团队 编辑丨具身智能之心 本文只做学术分享，如有侵权，联系删文 >> 点击进入→ 具身智能之心 技术交流群 更多干货，欢迎加入国内首个具身智能全栈学习社区 ： 具身智能之心知识星球 (戳我) ， 这里包含所有你想要的。 11月17号！Physical Intelligence团队正式发布 ，从经验中学习的VLA。 项目链接：https://www.pi.website/blog/pistar06 论文链接：https://www.pi.website/download/pistar06.pdf VLA模型如何通过强化学习在现实部署中实现自我改进？ 提出了一种通用方法RECAP：基于经验与校正的优势条件策略强化学习，该方法通过优势条件机制 实现VLA模型的强化学习训练。 该方法将异构数据整合到自我改进过程中，包括演示数据、在线收集数据以及在自主执行期间专家远程干预数据。RECAP方法首先通过离线强化学习预训练通用型 VLA模型（记为 ），该模型随后可通过机器人现场数据收集实现下游任务的专业化性能提升。 实验表明 ...

贝佐斯亲身下场物理AI了，亲自担任CEO的那种。 纽约时报消息，这名前世界首富创立了一家新公司并亲自担任联席CEO。 而且资金实力雄厚，包括贝佐斯本人出资在内，该公司已获得62亿美元资金。 编辑丨 量子位 点击下方 卡片 ，关注" 具身智能之心 "公众号 >> 点击进入→ 具身 智能之心 技术交流群 更多干货，欢迎加入国内首个具身智能全栈学习社区： 具身智能之心知识星球（戳我） ，这里包含所有你想要的! 这也是贝佐斯卸任亚马逊CEO之后，首次担任正式的运营职务。 贝佐斯入局物理AI 去年，贝佐斯投资了具身大牛 Sergey Levine 创立的顶尖AI机器人公司Physical Intelligence，现在他又亲自下场创立了 Project Prometheus ，进军物理AI。 有贝佐斯的下场，这个公司一创立就资金雄厚，获得了62亿美元，约合人民币440亿。 员工规模也达到了上百人，其中包括从OpenAI、DeepMind等顶级人工智能公司挖来的研究人员。 Project Prometheus的研究项目包括将人工智能应用于机器人、药物设计和科学发现等物理任务，明确将重点放在计算机、汽车、航空航天 等高科技 ...

图为2025年美国拉斯维加斯消费电子展Old World Labs展区的机器人展示。（ 新华社发） 近期，美股中人工智能（AI）概念股波动加剧，华尔街关于"人工智能泡沫"的警告声日益响亮。一边是 英伟达等AI概念股市值屡创新高、互联网巨头不断加码算力建设；一边是"大空头"公开做空AI明星股， 软银集团、桥水基金、花旗等机构清仓或大幅减持英伟达股票。进与退的交错之间，显露出资本市场对 AI估值的分歧正在加大，对行情过热的担忧情绪正逐渐升温。 当前市场对美国"人工智能泡沫"的质疑主要集中体现在以下几个方面。 一是概念股估值过高透支业绩。2022年11月ChatGPT大模型出现被视为推动AI产业进入新增长阶段的关 键节点，自其问世以来，英伟达公司股价已累计上涨约11倍，同期，标普500指数上涨约70%。今年10 月29日，英伟达公司市值首次突破5万亿美元，超过德国全年国内生产总值（GDP）规模。在公布获得 美国开放人工智能研究中心（OpenAI）的3000亿美元订单后，云服务商甲骨文股价一度飙升41%，但 最新财报显示，AI云服务毛利率仅14%，租赁英伟达芯片单季亏损1亿美元。 二是基础设施投资过度。以OpenAI ...

据港交所11月18日披露，诺比侃人工智能科技(成都)股份有限公司(简称：诺比侃)递表港交所主板，中金公司（601995）为其独家保荐人。这是该公司第三 次递表港交所。 于往绩记录期间，该公司的收入来自轨道交通、电力及城市治理业务。该公司亦已扩展至其他垂直行业领域，包括城市交通、机场及化工，截至最后实际可 行日期(2025年11月9日)，该等新业务的商业化处于不同业务及╱或产品开发阶段。 通过行业经验的累积和持续研发投入，该公司搭建了整合数据采集、预处理、数据标注及分析以及模型训练及优化等功能模块的底层AI技术平台-NBK- INTARI人工智能平台。根据灼识咨询，该公司的NBK-INTARI人工智能平台在大数据处理、深度学习及基础算法领域处于中国AI技术底层平台的领先地 位。 该公司针对不同业务场景开发不同的AI行业模型，并不断对算法模型及产品进行优化升级。AI行业模型是指针对特定业务场景定制的算法模块的集合。算 法模块指一款AI系统中执行一项特殊任务或一系列任务的独立组件或指令集。该等算法模块旨在处理与其目标业务场景相关的数据处理、分析或决策的特 定方面。该公司的AI行业模型是多模态的，涵盖视觉模态、仿真模 ...

Group 1 - Xiaomi will release an upgraded version of its end-to-end assisted driving system, HAD Enhanced Version, at the Guangzhou Auto Show on November 21 [5] - Dazhong City, a state-owned real estate enterprise, has announced its plan to delist from the Hong Kong Stock Exchange, with the delisting expected to take effect on November 27 [3] - Wahaha's general manager represented the company at the 2025 sales meeting, while the founder, Zong Fuli, was absent [4] Group 2 - Google has launched its latest AI model, Gemini 3, which aims to compete with OpenAI's ChatGPT, boasting improved reasoning and multimodal capabilities [2] - Apple's designer Abidur Chowdhury has reportedly left the company to join an AI startup, causing a stir within Apple's design team [6] - The tax authorities are cracking down on tax evasion practices involving income splitting, with several cases exposed this year [11] Group 3 - Meta's Chief Revenue Officer, John Hegeman, announced his departure to start his own company, marking a significant leadership change for the social media giant [7] - Meta won a federal antitrust lawsuit regarding its acquisitions of Instagram and WhatsApp, with the court ruling that the transactions did not violate antitrust laws [7] - The U.S. stock market saw a collective decline, with major indices falling, and many large tech stocks also experienced drops [8] Group 4 - Amazon is set to issue bonds for the first time in three years, raising $15 billion to support AI development [18] - "Material Science Source Map" has completed a multi-million angel round of financing, which will focus on core technology development [19] - Weibo reported a 5% year-on-year decline in total revenue for Q3, amounting to approximately 3.156 billion yuan [20]

智东西11月19日报道，今天凌晨，谷歌最强推理模型Gemini 3终于亮相，一个模型囊括了原生多模态、推理、Agent多种能力。 谷歌DeepMind研究团队称，这是全球最先进的多模态理解模型、谷歌最强大的Agent编程和氛围编程模型，能呈现更丰富的可视化效果和更深度的交互体 验，且完全构建于最先进的推理技术基础之上。 该模型基于谷歌TPU进行训练，支持100万个token的上下文窗口，适用于需要以下功能的应用：Agent、高级编程、长上下文、多模态理解、算法开发。 OpenAI联合创始人、CEO萨姆·阿尔特曼（Sam Altman）和xAI创始人、CEO埃隆·马斯克（Elon Musk）纷纷向谷歌发来"贺信"。阿尔特曼发推评 价"Gemini 3看起来很不错"，谷歌CEO桑达尔·皮查伊（Sundar Pichai）用了个表情包回复。 马斯克转发了谷歌DeepMind CEO戴密斯·哈萨比斯（Demis Hassabis）的推文称"干得不错"。 今天起，谷歌将在以下平台部署Gemini 3： 适用于Gemini应用的所有用户，以及在搜索的AI模式中使用Google AI Pro和Ultra订阅服务的用户；适用 ...

Core Viewpoint - The report from China International Capital Corporation (CICC) maintains a positive outlook on sectors related to AI, electrification, and finance, highlighting sustained demand and investment opportunities in these areas [1] AI Sector - CICC analysts predict that the high demand for overseas computing power will continue until the end of 2026, with tight capacity expected in the electricity and storage sectors during the same period [1] - The growth in overseas electricity consumption is driven by AI infrastructure and industrial chain restructuring, indicating a persistent narrative of electricity shortages that will benefit the power operation sector [1] Electrification - Electrification is expected to benefit from energy transition, infrastructure projects, and investments in data centers, contributing to the overall growth in the sector [1] Consumer Sector - CICC expresses a cautious stance on the consumer sector, forecasting a potential slowdown in consumption growth in the U.S. by 2026, while non-U.S. regions may see marginal recovery but remain slow overall [1] - In the context of ongoing fundamental differentiation, the report suggests focusing on sectors with reasonable valuations and positive potential catalysts, such as marginal changes in pharmaceuticals and automotive industries [1]

| 市场 | 名称 | 现价 | 涨跌 | 涨跌幅 | | --- | --- | --- | --- | --- | | 欧美 | 道琼斯工业平均 | 46091.74 | -498.5 | -1.07% | | | 纳斯达克 | 22432.85 | -275.22 | -1.21% | | | 标普500 | 6617.32 | -55.09 | -0.83% | | | 欧洲斯托克50 | 5533.53 | -107.41 | -1.90% | | | 英国富时100 | 9552.3 | -123.13 | -1.27% | | | 法国CAC40 | 7967.93 > | 5151.09 www.gogudata.com | -1.86% | | | 德国DAX | 23173.05 | -417.47 | -1.77% | | | 俄罗斯RTS | 996.14 | 21.28 | 2.18% | | | 上证指数 | 3939.81 | -32.22 | -0.81% | | | 深证成指 | 13080.49 | -121.51 | -0.92% | | | 创业板指 | 3069.22 ...