联想集团董事长兼CEO杨元庆表示，随着agentic AI推动推理工作负载呈指数级增长，成本控制和单 token性能将变得至关重要。通过将NVIDIA AI Enterprise软件与联想全栈混合AI平台和服务结合，联想 能够帮助客户以更高效率、更低单token成本、更快的投产速度，来扩展AI应用规模。 新华财经北京3月17日电（记者沈寅飞）记者从联想集团获悉，当地时间3月16日，在美国加州举行的 GTC 2026大会上，联想集团与英伟达联合发布新一代联想Hybrid AI Advantage （混合式AI优势集）解 决方案，旨在加速AI落地、缩短首token时间（TTFT），并在个人、企业和云环境中带来可量化的商业 成果。 据介绍，此次联想的新一代的混合式AI解决方案，包括从终端扩展到数据中心，再扩展至吉瓦级AI云 部署，从而在全球范围内赋能实时决策、运营效率提升和智能自动化。大会上，联想集团还正式成为 NVIDIA Vera Rubin NVL72的全球首发合作伙伴。较上一代相比，Vera Rubin系统吞吐量提升高达10 倍，单token成本下降至前一代的1/10，将极大实现更快部署、更低成本。 （文章 ...

英伟达CEO黄仁勋强调，"自动驾驶汽车革命已然降临——这是首个突破万亿美元规模的机器人产业。 通过英伟达Hyperion平台与Alpamayo开放式推理模型，我们赋予车辆三大核心能力：精准感知周边环 境、智能推理复杂场景、安全执行驾驶决策——这为可扩展的L4级自动驾驶奠定了技术基石。"(格隆 汇) 优步(UBER.US)盘前涨超2%，报76.19美元。消息面上，英伟达和优步宣布，计划将于2027年上半年首 先在美国洛杉矶及旧金山湾区部署自动驾驶出租车车队。随着技术架构的成熟，该网络预计在2028年迅 速扩张至全球四大洲的28个核心城市，目标是最终在优步平台上部署超过10万辆搭载英伟达尖端AI技 术的L4级自动驾驶车辆。 ...

Group 1: Nvidia - Nvidia develops GPUs, CPUs, and high-performance networking equipment, holding over 80% market share in AI accelerators [5][6] - The company reported a 73% increase in revenue to $68 billion and an 82% increase in non-GAAP earnings to $1.62 per diluted share for the fourth quarter of fiscal 2026 [8] - Nvidia's median target price among 69 analysts is $265 per share, implying a 47% upside from its current price of $180 [9] - The stock trades at 38 times adjusted earnings, which is considered cheap given the expected 39% annual earnings growth through fiscal 2029 [10] Group 2: Micron Technology - Micron develops memory and storage solutions, including DRAM and NAND flash memory products, and is the third largest supplier in these categories [11][12] - The company reported a 56% increase in revenue to $13.6 billion and a 167% increase in non-GAAP net income to $4.78 per diluted share in the first quarter [15] - Micron's median target price among 49 analysts is $450 per share, implying a 6% upside from its current price of $426 [9] - The stock trades at 38 times adjusted earnings, but Wall Street expects earnings to decline sharply after fiscal 2027, making the current valuation appear expensive [16]

Core Viewpoint - Nvidia is set to launch high-efficiency AI chips in the second half of 2026, with procurement orders reaching $1 trillion by 2027, indicating strong demand for AI technology and automation solutions [2][5]. Group 1: Product Launch and Efficiency - Nvidia will introduce new AI chips in the second half of 2026, which can achieve power efficiency up to 35 times higher than previous models [2]. - The new chips are essential for the development of AI agents that facilitate business automation [2]. Group 2: Market Demand and Orders - The demand for AI chips, particularly for the latest models "Rubin" and "Blackwell," is expanding, with procurement orders projected to reach $1 trillion by 2027 [5]. - CEO Jensen Huang emphasized that AI has reached a turning point where it can effectively handle productivity tasks [5].

Samsung Electronics shares rose sharply on Tuesday after fresh signals from Nvidia strengthened expectations of a turnaround in its chipmaking business. The stock climbed as much as 5% in early trade ... ...

Core Insights - NVIDIA's GTC 2026 showcased a significant shift in the AI industry, with CEO Jensen Huang emphasizing the company's projected revenue of at least $1 trillion by 2027, up from $500 billion last year [4][12][36] Group 1: Event Overview - The GTC 2026 featured 450 sponsoring companies, 1,000 technical sessions, 2,000 speakers, and 110 robots, resembling an annual pilgrimage for the AI industry rather than a typical tech conference [1] - Huang's keynote included a historical overview of NVIDIA's technology evolution, from the GeForce graphics card to the current AI advancements [4][36] Group 2: Token Economy - Huang introduced a comprehensive business model based on token economics, highlighting different pricing tiers for token usage, ranging from free access to $150 per million tokens for advanced tasks [9][12] - The model emphasizes the importance of token throughput and generation rates, which are critical for AI performance [9][12] Group 3: Technological Advancements - The new Vera Rubin AI computing system is described as NVIDIA's most complex, featuring seven types of chips and achieving 3.6 exaflops of computing power [15][20] - Innovations include a 100% liquid cooling solution and the introduction of Co-Packaged Optics (CPO) technology, which enhances data transmission efficiency [21][24][25] Group 4: Integration of Groq - NVIDIA's acquisition of Groq aims to enhance AI inference capabilities, with Groq's architecture designed for high-speed processing, complementing NVIDIA's GPU offerings [29][31] - The integration strategy involves separating inference tasks between Groq and Vera Rubin to optimize performance [33][35] Group 5: Future Developments - Huang announced the upcoming Feynman architecture, which will introduce new GPUs, LPUs, and CPUs, emphasizing the need for higher capacity and bandwidth in future computing solutions [44][47] - The company is also exploring the concept of "space computing" with plans for a data center in space, addressing unique challenges such as heat dissipation in a zero-gravity environment [49][50]

Core Insights - Silicon Motion Technology Corporation is showcasing its differentiated enterprise SSD controllers and PCIe NVMe boot SSD solutions at NVIDIA GTC 2026, aimed at meeting the evolving requirements of the NVIDIA AI ecosystem [1][16][10] Group 1: Product Offerings - The company presents a vertically integrated storage solution that includes advanced SSD controller design, full firmware development, and compact Reference Design Kits (RDKs) for AI server deployments [3][11] - Silicon Motion's MonTitan® enterprise SSD controllers are designed to meet the demands of AI infrastructure, providing high capacity, deterministic latency, and predictable quality-of-service [4][6] - The enterprise PCIe NVMe BGA boot SSDs are engineered for compact integration, strong endurance, and long-term stability, essential for AI systems [8][9] Group 2: Market Trends - As AI models scale, inference architectures are extending into high-performance NAND storage tiers, making NAND storage a performance-critical layer that requires deterministic latency and quality-of-service differentiation [2][5] - The NVIDIA Inference Context Memory Storage (ICMS) initiative emphasizes the need for tight integration between intelligent data management software and high-performance storage hardware [5][4] Group 3: Technical Innovations - Silicon Motion's controllers utilize patented PerformaShape™ technology to optimize workload behavior, enhance isolation across concurrent workloads, and manage latency distribution [6][4] - The MonTitan® controllers support various storage requirements, including high bandwidth and IOPS for near-GPU acceleration, throughput consistency for AI compute nodes, and high-density storage for nearline SSD applications [7][6][8]

本文字数：949，阅读时长大约1.5分钟 作者 | 第一财经 钱童心 当地时间3月16日，瑞士制药巨头罗氏宣布，已经在美国和欧洲等多地部署了2176个英伟达AI计算 芯片，如此大规模算力的部署超过了本月早些时候礼来AI工厂的算力规模。 近期，礼来宣布AI制药工厂LillyPod已投入运行，这是全球首个完全由制药企业自主运营的AI制药工 厂。礼来称该工厂仅用四个月便完成组装，搭载超过1000个英伟达Blackwell Ultra GPU，将为科学 研究提供巨大算力。 罗氏也表示，最新的AI项目部署的超过2100个英伟达芯片将有助于扩展该公司人工智能的计算能 力，以支持药物和诊断学的发展。"这些硬件将加快研发业务工作，包括建模、数据分析和临床试验 过程。"罗氏方面称。 近年来，大型制药公司竞相加大对AI的投资力度，以缩短产品开发周期并降低费用，罗氏、礼来、 阿斯利康、辉瑞等巨头纷纷通过合作交易的方式购入具有前景的AI工具。 今年1月，礼来与英伟达宣布，将在未来五年内斥资10亿美元在旧金山湾区建立一个新的联合研究实 验室，以加速AI药物研发进程。阿斯利康也于今年早些时候首次购入一家人工智能公司Modella AI， ...

华尔街目光聚焦GTC大会之际，全球芯片巨头三星电子再出王炸。 当地时间3月16日，三星在2026年GTC大会上公开展出其下一代高带宽内存芯片HBM4E。 这也是三星第七代HBM技术的首次亮相。 与此同时，英伟达CEO黄仁勋透露，英伟达的新型人工智能芯片正在由三星生产。 受此消息提振，周二，三星电子股价大幅上涨，截止发稿，涨幅高达3.29%。 HBM4E重磅发布 具体来看，三星在本次GTC大会的展示核心，主要是现已实现量产的第六代HBM4和首次亮相的 HBM4E。 据悉，HBM4在带宽、功耗和热管理方面均较前代显著改进，现"已进入商业化阶段"。 三星表示，依托HBM4量产过程中积累的技术优势，公司正基于1c DRAM 制程工艺，加速推进下一代 HBM4E的研发进程。 根据官方披露的信息，HBM4E单引脚速率可达16Gbps，单堆栈带宽最高约4TB/s，主要面向下一代AI和 高性能计算系统。 韩国SK集团会长崔泰源预计，DRAM、NAND和 HBM等各类存储芯片的价格将持续上涨，涨势可能会 持续较长时间。 "全球内存芯片短缺的情况很可能会持续到2030年。" 为了满足各大科技企业的芯片订单，三星正着手在其得克萨斯 ...

2026 年 3 月 17 日 英伟达 NVDA US 本文内容由 Kylie Kwok (郭凯欣) 提供 CSIWM 个股点评 中信证券财富管理(香港) 产品及投资方案部 美国科技行业 电话：(852) 2237 9250 / 电邮：wminvestmentsolutions@citics.com.hk GTC 2026：一系列公告 摘要 中信证券财富管理与中信里昂研究观点一致。根据中信里昂研究在 2026 年 3 月 17 日发布的题为《GTC 2026: Announcements galore》的报告，英伟达 GTC 大会发布了一系列公告，预示着其显著的上行空间。公司关于"到 2027 年"计算收入达 1 万亿美元的预测高于分析同期数据中心的预测。此外，Groq LPU 将与 Rubin 机架级系统协同 工作，其宣称的 35 倍速度提升有望颠覆现有拥有成本计算模型。分析表示，这些公司公告中蕴含的规模与复杂性仍 需时间消化。 中信证券财富管理(香港) 免责声明请参考封底 1 万亿美元具体指什么？分析认为该指引涉及英伟达"到 2027 年"的计算收入预期，即从现在到 2027 年 （未来八个季度）。按此推 ...