美股频道更多独家策划、专家专栏，免费查阅>> 来源：环球市场播报 三星电子在英伟达 GTC 大会上首次展示了这款芯片，突显了其成为英伟达即将推出的 Vera Rubin 平台 关键合作伙伴的努力。三星还展示了 Hybrid Copper Bonding（混合铜键合），这是一种旨在克服传统热 压键合物理限制的新封装技术。HCB 技术可让新一代高带宽存储器（HBM）实现16 层及以上堆叠，同 时将热阻降低20% 以上。其他为英伟达 AI 基础设施打造的前沿技术还包括HBM4与SOCAMM2，目前 这两款产品均已量产。 责任编辑：栎树 ...

🚨 Jensen Huang: “Today, we are announcing four new partners for NVIDIA’s Robotaxi-Ready platform: BYD, Hyundai, Nissan, and Geely.”https://t.co/8h7pMHNeUY ...

SK海力士公司周二表示，将参加由英伟达公司在美国加州主办的年度技术大会，重点展示其在人工智 能（AI）行业内存芯片领域的领先地位。 SK海力士表示，该公司将以"聚焦AI内存"为主题，在周一（美国时间）开幕的英伟达GTC 2026大会上 设立展位，该大会将持续四天。 此次展览将展示SK海力士的主要产品，例如第六代高带宽内存（HBM4），如何应用于英伟达的AI平 台。 SK海力士表示，还将展示其内存解决方案，这些方案"旨在最大限度地减少数据瓶颈，并最大限度地提 高英伟达AI平台上的AI训练和推理性能"。 SK集团董事长崔泰元和SK海力士首席执行官郭鲁正也计划在活动期间会见包括英伟达在内的多家大型 企业的代表，以拓展业务合作关系。 该公司在一份声明中表示："随着人工智能技术的演进，内存不再仅仅是一个组件，而是成为决定整个 人工智能基础设施架构和性能的核心要素。" 该公司补充道："通过将其在内存领域的专业知识应用于从数据中心到设备端应用的整个人工智能领 域，该公司旨在与全球合作伙伴一起塑造人工智能的未来。" 责任编辑：于健 SF069 SK海力士公司周二表示，将参加由英伟达公司在美国加州主办的年度技术大会，重点展示其 ...

黄仁勋公布此预测后，英伟达股价刷新日高、涨近5%。去年10月末黄仁勋说，2025至2026年，Blackwell和Rubin相关数据中心业务 将合计实现5000亿美元收入。 华尔街见闻 在全球AI算力竞赛持续升温之际，英伟达CEO黄仁勋给市场带来一个更为激进的增长预期。 美东时间3月16日周一，黄仁勋在英伟达年度开发者大会GTC上表示，英伟达预计，旗下新一代AI加速芯片架构Blackwell与下一代 Rubin产品，到2027年底累计将创造至少1万亿美元收入。这一数字远超黄仁勋去年10月给出的5000亿美元销售预测，再次凸显AI基 础设施投资浪潮仍在快速扩张。 黄仁勋公布上述1万亿美元的预测后，周一美股午盘时段，英伟达股价涨幅迅速扩大、刷新日高，日内涨约4.8%，后很快回吐过半涨 幅，到收盘时涨不足2%。在此之前，截至上周五收盘，英伟达股价进入2026年以来已累跌超3%，市场对AI投资周期持续性的担忧一 度令股价承压。 AI算力需求持续爆发 过去两年，随着生成式AI快速普及，英伟达已经成为AI基础设施浪潮中最核心的硬件供应商。 Blackwell成当前核心增长引擎 Rubin瞄准下一轮算力升级 在产品路线图中 ...

在GTC 2026大会上，英伟达CEO黄仁勋将公司定位为"AI工厂"构建者，称"到2027年将看到至少1万亿美元的高确信度需 求"。他提出"Token工厂经济学"，强调每瓦性能是商业变现的核心。黄仁勋断言Agent（智能体）将终结传统SaaS模式，未 来"年薪+Token预算"将成为职场新标配。 华尔街见闻 开场致辞 主持人： 欢迎英伟达创始人兼首席执行官黄仁勋登台。 黄仁勋，创始人兼首席执行官： 欢迎来到GTC。这是一场技术大会，看到这么多人一大早就排队入场，非常高兴。 今天我们将围绕三大平台展开讨论：CUDA-X平台、系统平台，以及全新的AI工厂平台。当然，最重要的是生态系统。 首先，我要感谢我们的"赛前热身"主持人，他们表现出色：来自Conviction的Sarah Guo、红杉资本的Alfred Lin（英伟达的第 一位风险投资人），以及英伟达第一位重要机构投资人Gavin Baker。这三位对技术有着深刻的理解，在技术生态系统中拥有 广泛的影响力。此外，也感谢我亲自挑选的所有贵宾。 同时，我要感谢所有参会企业。英伟达作为平台公司，拥有技术、平台和丰富的生态系统。今天，涵盖百万亿美元产业的企 业几乎全 ...

在技术层面，NemoClaw的亮点在于其对OpenClaw平台的无缝优化。通过NVIDIA Agent Toolkit软件， NemoClaw实现了对OpenClaw的"一键式"适配。其内置的OpenShell环境提供了一个隔离的沙箱，让每个 人都能轻松打造专属的高性能安全 AI 助手。 凤凰网科技讯 3月17日，黄仁勋在GTC大会上宣布推出全新的NemoClaw软件栈，旨在为"小龙 虾"OpenClaw智能体平台提供"一键安装"体验。 在大会现场的主题演讲中，黄仁勋展示了NemoClaw的便捷性。他仅通过一条指令，就快速部署了英伟 达的Nemotron模型与新发布的OpenShell运行环境。黄仁勋将这一突破提升到了战略高度，他形象地比 喻道："正如Mac和Windows成为了个人电脑的操作系统，OpenClaw将成为'个人AI的操作系统'。这不 仅是产品的发布，更预示着软件行业一个新时代的开启。" 这一架构允许AI智能体在严格确保数据隐私和网络安全的前提下灵活调用模型。它既可以调用本地专 属系统运行的开源模型（如Nemotron），也可以通过创新的"隐私路由器"安全地接入云端的前沿大模 型，兼顾了性能与安 ...

Core Insights - Nvidia forecasts a revenue opportunity of at least $1 trillion for its AI chips by 2027, significantly up from a previous estimate of $500 billion through 2026 [4] - The company aims to strengthen its position in inference computing, which involves answering queries, as it faces increased competition from CPUs and custom processors [2][8] - CEO Jensen Huang emphasized the growing demand for AI infrastructure during the annual GTC developer conference, indicating a shift in focus from AI model training to real-time AI system deployment [3][6] Revenue and Market Position - Nvidia's new forecast reflects a robust demand for its AI infrastructure, despite investor concerns about growth sustainability [6][11] - The company has introduced a new central processor and an AI system based on technology from Groq, which it acquired for $17 billion [1][5] - Nvidia's Vera Rubin chips will manage the initial "prefill" stage of inference, while Groq's chips will handle the "decode" stage, showcasing a strategic approach to enhance its AI capabilities [7] Competitive Landscape - Companies like OpenAI, Anthropic, and Meta are transitioning from training AI models to serving a larger user base, increasing demand for CPUs, which are seen as viable alternatives to Nvidia's GPUs [9][13] - Huang noted that the standalone CPU business is expected to become a multi-billion-dollar segment for Nvidia [10] - The company is also targeting the autonomous AI agent market with NemoClaw, which integrates privacy and safety controls into AI tools [14] Future Developments - Nvidia's Feynman architecture is anticipated to launch in 2028, following the Rubin Ultra chips, indicating a long-term vision for its product roadmap [13] - The announcements made at the conference have elevated discussions around AI infrastructure, suggesting a shift in how the industry approaches AI deployment [14][15]

Nvidia wants to simplify the creation and management of your own AI agent. The company unwrapped NemoClaw on Monday during its GTC Conference keynote, a reference stack for the OpenClaw agent platform. NemoClaw is Nvidia's specialized infrastructure layer that makes it easier to set up and run AI agents with an added security layer, according to the company. OpenClaw has taken the AI world by storm in recent months, with the term "claw" itself becoming the latest in AI lingo. CEO Jensen Huang described Open ...

Nvidia's GTC conference on Monday gave the world the latest from the AI chip behemoth, including a new AI-powered upscaling software for gaming and a new Vera CPU to supercharge the future of AI agents, physical AI and more. CEO Jensen Huang spelled out Nvidia's blueprint for remaining the dominant force in the AI industry, focusing on its data center platform, Vera Rubin, as he highlighted the trend toward agentic AI. If you missed the keynote on Monday, here are the three big announcements you need to kno ...

Core Insights - EXL has enhanced its agentic AI platform, EXLerate.ai™, with support from NVIDIA AI Enterprise, aiming to deliver secure and compliant AI solutions for regulated industries [1][3] - The collaboration with NVIDIA allows EXL to provide a robust foundation for enterprises in sectors like insurance, healthcare, and banking to scale their AI initiatives [3][7] Group 1: EXLerate.ai Platform Features - EXLerate.ai now includes over 250 pre-built AI agents and accelerators, which can reduce development costs by approximately 40% and decrease the time from prototype to production by up to 50% [2] - The platform features Agent Studio, a no-code, fully autonomous agent-building tool that simplifies the creation of secure workflows for IT and business users [4] - The AI Governance Hub within EXLerate.ai integrates advanced AI models and offers over 40 frameworks, enabling enterprises to manage risk and ensure compliance effectively [5] Group 2: Performance and Impact - EXLdata.ai, another platform from EXL, utilizes NVIDIA CUDA-X Data Science for GPU-accelerated data processing, achieving a 30-50% faster time-to-model while enhancing accuracy and compliance [6] - The collaboration with NVIDIA is expected to accelerate the development of solutions while ensuring reliability in highly regulated environments, supporting over 2,000 AI-powered workflows for more than 800 global clients [7]