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"。 这些动作表明：AI行业的竞争正从单纯的算力比拼，转向对单位面积带宽的极致追求——这与英伟达此前"大模型推理90%的延迟源于数据搬运，导 致算力利用率常低于30%"的结论不谋而合。 然而，当英伟达选择通过 ...

来源丨 量子位（ID：QbitAI） 作者丨西风 闻乐 刚刚，英伟达CEO黄仁勋穿着鳄鱼皮夹克，在全球最大消费电子展 CES 2026 上发布AI新品。 这是五年来，英伟达首次来到CES却没有发游戏显卡，态度很明确：全力 搞AI。 全力搞出来的结果也让围观群众直呼：竞争对手如何追上英伟达？ 下一代Rubin架构GPU 推 理、训练性能分 别是 Blackwell GB 200的5倍和3.5倍 （NVFP4数据格 式）。 除此之外，老黄还带来了五大领域的全新发布，包括： 面向Agentic AI的 NVIDIA Nemotron 模型家族 面向物理AI的 NVIDIA Cosmos 平台 面向自动驾驶开发的全新 NVIDIA Alpamayo 模型家族 同时，英伟达宣布持续向社区 开 源训 练框架 以 及 多模 态数据 集 。其中数据集包括10万亿语言 训练token、50万条机器人轨迹数据、45.5万个蛋白质结构、100TB车辆传感器数据。 这次的核心主题，直指 物理AI 。 用网友的话来说： 这是英伟达将护城河从芯片层进一步拓展到全栈平台层（模型+数据+工具）的体现，通过这种方式可 以持续拉动更多GPU与基 ...

Core Viewpoint - Nvidia's Q3 FY26 financial results exceeded Wall Street expectations, showcasing significant growth in revenue and net profit driven by strong demand for AI infrastructure and GPU sales [1][2]. Financial Performance - Nvidia reported revenue of $57.006 billion, a year-over-year increase of 62% and a quarter-over-quarter increase of 22% [1][9]. - Non-GAAP net income reached $31.767 billion, reflecting a 59% year-over-year growth and a 23% quarter-over-quarter increase [9]. - The company achieved a non-GAAP gross margin of 73.6%, up 0.9 percentage points from the previous quarter but down 1.4 percentage points year-over-year [8][9]. Revenue Breakdown - The data center segment generated $51.215 billion, a 66% increase year-over-year and a 25% increase quarter-over-quarter [7][9]. - The compute segment contributed $43.028 billion, with a 56% year-over-year growth and a 27% quarter-over-quarter increase [7][9]. - Networking revenue surged by 162% year-over-year, reaching $8.187 billion [7][9]. - Gaming and professional visualization segments also saw growth, with gaming revenue at $4.265 billion (30% year-over-year) and professional visualization at $760 million (56% year-over-year) [7][9]. Market Dynamics - Nvidia's CEO highlighted three major platform transitions: the shift from CPU to GPU computing, the rise of generative AI applications, and the emergence of Agentic AI [1][10]. - The demand for AI infrastructure is outpacing Nvidia's expectations, with major cloud service providers experiencing sold-out capacities [2][10]. - Nvidia's partnership with Anthropic, involving a combined investment of $15 billion, underscores the company's strategic positioning in the AI market [12]. Future Outlook - Nvidia anticipates revenue of $65 billion for Q4 FY26, with a projected non-GAAP gross margin of 75% [9][14]. - The company expects to benefit from increased capital expenditures in the AI infrastructure sector, with top cloud providers' spending projected to reach $600 billion, up $200 billion from earlier estimates [14].