如果您希望可以时常见面，欢迎标星收藏哦~ RISC-V 领 军 企 业 SiFive 加 入 了 越 来 越 多 的 芯 片 公 司 行 列 ， 全 力 支 持 英 伟 达 专 有 的 NVLink Fusion 互连技术，此举令人怀疑竞争对手 UALink 互连技术的可行性。 我们可能永远不会知道是否有人使用 SiFive 的 NVLink 兼容性功能，因为该公司的客户没有披露 他们已获得哪些 IP 许可的习惯（除了一些个别情况）。 尽管如此，该公司告诉我们，它拥有"多个数据中心设计许可，其中一些已经完成流片"。但目前它 还不准备宣布任何新的许可。 SiFive公司基于RISC-V指令集架构（ISA）设计CPU内核和处理器，应用范围涵盖从边缘计算到 数据中心等各种场景。该公司将其设计授权给客户，客户可将其用于自己的芯片中，这种模式与 Arm的商业模式类似。 随着今天发布的公告，SiFive 的设计现在将包含对 NVLink 的支持。 NVLink 使英伟达的客户能够将一整架 CPU 和 GPU 抽象为一个统一的加速器。该技术可提供高 达 3.6 TB/s 的芯片间带宽。 直到最近，Nvidia 一直将 NV ...

在AI算力赛道，英伟达凭借Hopper、Blackwell、Rubin等架构GPU，早已在AI训练领域建立起了难以撼动的技术壁垒与行业地位。但随着 即时AI场景需求爆发，传统GPU在面对低批处理、高频交互推理任务中的延迟短板愈发凸显。 为破解这一痛点，英伟达重磅出击，斥资200亿美元收购Groq核心技术，抢跑AI推理市场。 这一金额不仅创下英伟达历史最大手笔交易、刷新了推理芯片领域的估值纪录，更鲜明地昭示着英伟达从"算力霸主"向"推理之王"转型的意志。 紧随这一动作，据技术博主AGF消息进一步披露，英伟达计划在2028年推出新一代Feynman架构GPU——采用台积电A16先进制程与SoIC 3D堆叠 技术，核心目的正是为了在GPU内部深度集成Groq那套专为推理加速而生的LPU（语言处理单元），相当于给GPU加装了一个专门处理语言类推理 任务的专属引擎，直指AI推理性能中长期存在的"带宽墙"与"延迟瓶颈"。 回看中国市场，AI浪潮推动下，国产大模型多点突破、强势崛起，本土AI芯片企业集体爆发并密集冲击IPO，资本热度居高不下。 然而，当英伟达选择通过Feynman架构来补齐推理短板时，就意味着谁能率先解决" ...

事关铜需求，英伟达大幅下调。 据最新消息，英伟达将其一篇技术论文中有关数据中心铜需求量数据进行了修正，将每吉瓦机架铜母线用量 从"50万吨"大幅下调至200吨。调整后，市场对铜未来需求的预期或将下调。市场层面，今日，伦铜价格直线 跳水，日内一度大跌近3%。 有分析指出，英伟达的大幅修正或许意味着，市场此前关于数据中心将导致铜陷入严重短缺的预期过于激进。 高盛分析师们警告称，铜价格上涨的大部分或许已经过去，铜价未来越来越容易出现回调。 英伟达大幅下调 北京时间1月16日，伦敦金属交易所(LME)铜期货价格直线跳水，日内一度大跌2.84%，截至17:10，跌幅为 1.82%，报12868美元/吨。周四，铜期货价格从历史高位回落，伦交所价格大跌超2%。 消息面上，英伟达将其一篇技术论文中有关数据中心铜需求量数据进行了改正，将传统数据中心一吉瓦机架的 铜母线用量修正为200吨，仅为原论文表述"50万吨"的0.04%。这意味着，数据中心的铜需求预期将被大幅调 低。 值得一提的是，英伟达官网去年刊登的一份技术论文指出，一个容量在一吉瓦的传统数据中心机架，需要高达 50万吨铜母线，这直接引起了外界对未来铜的供需紧张形势的担 ...

Core Insights - The article discusses the potential investment opportunities in lesser-known companies that are crucial to the AI infrastructure, rather than just focusing on high-profile AI companies like Nvidia and Super Micro Computer [1][3]. Group 1: AI Market Overview - The AI market has seen significant growth, with early leaders like Super Micro Computer and Nvidia posting gains of 231% and 1,066% respectively over three years, while the S&P 500 index gained 77% during the same period [2]. - Despite the impressive gains of these companies, there is a belief that the next wave of investment opportunities may lie in companies that provide the infrastructure necessary for AI [3]. Group 2: Hewlett Packard Enterprise (HPE) - Hewlett Packard Enterprise (HPE) is positioned as a key player in the AI infrastructure, assembling AI supercomputers and selling them to various sectors, including enterprises and governments [6][7]. - HPE's stock is currently valued at a modest 8.0 times forward earnings and 0.9 times trailing sales, making it an attractive investment compared to higher-valued chip designers [8]. Group 3: Schneider Electric - Schneider Electric is highlighted as a significant but underappreciated player in the AI data center space, providing essential power distribution and cooling systems for AI facilities globally [10][11]. - The company has a market capitalization of approximately $155 billion and annual sales around $42 billion, with its products being utilized by the largest cloud computing and AI software companies [12].

U.S. Stock Market Factors to Watch Out on FridayThe frenzy around AI has sent Nvidia and other superstar stocks to dizzying heights, stirring criticism that their prices had shot too high. Nvidia rose 2.1 per cent on Thursday after TSMC’s Chief Financial Officer Wendell Huang said it’s seeing “continued strong demand” in an encouraging signal for the entire AI industry.TSMC’s stock that trades in the United States rose 4.4 per cent on Thursday. Wall Street steadied on Thursday as stocks related to artificia ...

Core Insights - The stock market experienced significant gains during President Trump's administration, with the S&P 500 closing up 16% in 2025, marking the third year of a bull market [1] - Trump's tax policy, particularly the Tax Cuts and Jobs Act (TCJA), has been a major driver of corporate investment trends, leading to a surge in stock buybacks [3][8] - The TCJA reduced the corporate income tax rate from 35% to 21%, the lowest since 1939, which has incentivized companies to invest in share repurchases [9][10] Stock Market Performance - The S&P 500 index rose by 16% in 2025, following a turbulent period related to Trump's trade policies [1] - During Trump's first term, major indices like the Dow Jones and Nasdaq saw substantial increases, with the S&P 500 rising by 70% [2] Impact of Tax Policy - The TCJA has led to a significant increase in corporate buybacks, with S&P 500 companies on track to repurchase an estimated $1.02 trillion in shares for 2025 [12] - Prior to the TCJA, quarterly buyback activity for S&P 500 stocks was between $100 billion and $150 billion, which surged to between $200 billion and $250 billion post-TCJA [13] Corporate Buybacks - In Q3 2025, S&P 500 companies bought back $249 billion worth of their own stock, down from a record $293.5 billion in Q1 2025 [12] - Apple has been a leader in share repurchases, buying back over $816 billion since 2013, with $90.7 billion spent in fiscal 2025 [17] - Alphabet ranks second in buybacks among S&P 500 companies, having repurchased $342.4 billion over the last decade [18] - Nvidia has also engaged in significant buybacks, totaling $115.1 billion over the last decade, with a recent annual buyback approaching $52 billion [21] Conclusion - The combination of Trump's tax policies and the resulting corporate strategies has led to a robust environment for stock buybacks, significantly impacting the financial landscape of major companies like Apple, Alphabet, and Nvidia [22]

海通国际发布研报称，英伟达(NVDA.US)与礼来(LLY.US)共建AI+制药实验室，双方合作意味着AI将从 工具层升级为药物研发的核心生产力，推动制药模式由"经验驱动"向"数据与算法驱动"转型。 Tempus(TEMP.US)业绩高增长共同验证AI正在从概念走向制药与医疗核心生产力。从技术面上看，AI 技术近年取得重大突破，AI在药物发现、临床前研究和临床试验等全流程都能大幅提高流程效率。当 前AI+医药进入产业化加速期，在人工智能与医药双领域共振下，看好AI+医疗发展前景。 海通国际主要观点如下： AI+制药：人工智能赋能制药全产业链 英伟达与礼来宣布成立首个专注于将人工智能应用于药物研发关键挑战的AI+制药联合创新实验室。该 实验室将汇集礼来在药物发现、开发与制造方面的全球领先专业知识及英伟达在人工智能、加速计算和 AI基础设施方面的技术优势。双方计划在未来五年投入最多10亿美元用于人才、基础设施与算力资源 建设，这标志着AI正式进入全球头部药企核心研发体系。海通国际认为，英伟达与礼来的合作意味着 AI将从工具层升级为药物研发的核心生产力，推动制药模式由"经验驱动"向"数据与算法驱动"转型。 AI+医 ...

Group 1 - Nvidia stock has remained relatively stable over the past three months, indicating a period of consolidation [1] - RBC Capital Markets analyst Srini Pajjuri is optimistic about a potential new rally for Nvidia shares [1]

Core Viewpoint - Nvidia is expected to maintain its position as the leading provider of AI infrastructure for the foreseeable future, driven by its dominant market share and robust product offerings [3][8]. Group 1: Market Position and Performance - Nvidia holds nearly 90% market share in the data center accelerator market, which is crucial for AI workloads [3]. - The company reported a 62% increase in revenue to $57 billion, with non-GAAP net income rising 60% to $1.30 per diluted share [10]. - Nvidia's stock has surged 1,150% since January 2023, yet analysts suggest the current valuation remains attractive at 45 times earnings [10][13]. Group 2: Product Offerings and Innovations - Nvidia's GPUs are complemented by CPUs and networking gear, along with a comprehensive ecosystem of software development tools, enhancing its competitive edge [4]. - The upcoming Vera Rubin superchip, set to launch in the second half of 2026, will significantly enhance performance, with Vera CPUs offering double the performance of previous models and Rubin GPUs running AI workloads much faster [11]. Group 3: Growth Opportunities - The demand for generative AI is a key growth driver, with Nvidia providing essential hardware and software for AI model development [5]. - Data center GPU sales are projected to grow at 36% annually through 2033, while robotaxi chip sales are expected to increase at 74% annually through 2030 [8][9]. - Nvidia is also positioned to benefit from the easing of export restrictions to China, potentially unlocking a significant market for its products [12].

中美在生成式AI和物理AI领域进行开发竞争 特朗普15日针对英伟达在台湾制造的H200等半导体加征25%关税，但有条件地允许对华出口。出口对 象被认为的是在中国大陆开展业务的西方国家企业。据称目前中国大陆尚未允许进口H200…… 特朗普1月14日签署了关于加征关税的行政命令。这是基于《贸易扩展法》第232条的行业关税。美国国 内使用的进口商品不征收关税。设想适用关税的是英伟达为了向中国大陆出口在台湾制造的H200而暂 时将其运到美国等情况。 特朗普2025年12月表示，作为允许H200等对华出口的交换，将征收25%的"上缴金"。此次的加征关税实 际上相当于上缴金。 在加征关税之前，美国商务部1月13日提出了将AI半导体对华出口放宽为许可制的新规。 H200和美国AMD的"MI325X"向中国供货时，必须在美国国内接受第三方机构的检测，因此可能会被征 收额外关税。 美国要求英伟达等对中国的出口量不得超过对美国出货量的50%。为了优先美国国内的需求，还要求把 美国国内剩余的部分用于对华出口。 在此基础上，美国商务部工业和安全局（BIS）将逐案审查可否出口。出口对象被认为的是在中国开展 业务的西方国家的企业。原则上 ...