英伟达营收因AI需求增长逾一倍
英伟达连续第13个季度创下营收纪录，人工智能基础设施需求持续激增。首席执行官黄仁勋表示，营收可能进一步增长。
原文深度解读
本文报道英伟达（Nvidia）2027财年第二季度（截至7月26日）业绩，营收连续第13个季度创纪录，同比增长106%至962.2亿美元，主要受AI基础设施需求驱动。数据中心业务营收同比增长117%至890亿美元，公司已开始量产Vera Rubin NVL72系统，并供应给CoreWeave、Google Cloud、Microsoft Azure和Oracle Cloud。CEO黄仁勋预计下一年营收可能增长约70%。此外，英伟达获得SpaceX子公司SpaceXAI作为其首款CPU Vera的外部客户，并发布RTX Spark、DGX Station和Alpamayo 2 Super等产品。公司预测第三季度营收1080亿美元，毛利率74%。尽管业绩强劲，股价盘后下跌1.5%至209美元，后反弹至约218美元。
- 英伟达2027财年第二季度营收962.2亿美元，同比增长106%，连续第13个季度创纪录。
- 调整后每股收益2.22美元，毛利率75%，较去年同期上升2.2个百分点，较上季度上升0.1个百分点，略高于公司此前指引的74.9%。
- 分析师预期营收921.7亿美元，调整后EPS 2.10美元；实际营收超出约4.4%，EPS超出约5.7%。
- 数据中心业务营收同比增长117%至890亿美元；Vera Rubin NVL72已开始量产，并供应给CoreWeave、Google Cloud、Microsoft Azure和Oracle Cloud。
- Vera Rubin NVL72每瓦吞吐量是上一代Blackwell平台的30倍，每token成本降至约1/35。
- 英伟达预测第三季度营收1080亿美元（中点），高于华尔街预期1041.9亿美元约3.7%；毛利率预测74%（中点），该展望不包括中国数据中心营收。
供应链影响
- AI基础设施需求持续增长可能推动对英伟达GPU、CPU及相关系统的需求，进而影响上游半导体制造、封装和测试产能。
- Vera Rubin NVL72的量产和交付可能增加对先进制程、HBM内存和电源管理等组件的需求，具体取决于供应链产能和良率。
- SpaceXAI采用Vera CPU可能扩大英伟达在CPU市场的份额，可能影响x86 CPU供应商的市场格局，但取决于实际部署规模。
- 英伟达对中国数据中心营收的排除可能影响其全球供应链布局，可能促使中国客户寻求替代方案，具体取决于地缘政治和出口管制政策。
- 毛利率预测略降至74%可能反映成本上升或产品组合变化，可能影响供应链议价和成本控制，但取决于实际执行。
系统解读，仅作为判断线索，不构成备货、出货、涨价或投资建议。
(Photo: Nvidia)
英伟达连续第13个季度创下营收纪录，人工智能（AI）基础设施需求持续激增。首席执行官黄仁勋表示，营收明年可能增长约70%，这凸显了他对AI市场持续扩张的信心。
英伟达于8月26日表示，截至7月26日的2027财年第二季度营收同比增长106%至962.2亿美元。调整后每股收益（EPS）为2.22美元，毛利率为75%。毛利率较去年同期上升2.2个百分点，较上季度微升0.1个百分点，略高于公司此前指引的74.9%。
业绩也超出了华尔街预期。分析师预计第二季度营收为921.7亿美元，调整后EPS为2.10美元。实际营收比预期高出约4.4%，EPS超出预期约5.7%。
数据中心业务推动了本季度业绩。数据中心营收同比增长117%至890亿美元。英伟达上个月开始量产其Vera Rubin NVL72系统，并向CoreWeave、Google Cloud、Microsoft Azure和Oracle Cloud供应这些系统。Vera Rubin NVL72每瓦吞吐量比上一代Blackwell平台高出多达30倍，同时将每token成本降至约1/35。
“算力就是营收，”黄仁勋表示，并补充说AI基础设施建设正在全速推进，Vera Rubin就是为了支持这一时刻而打造的。
英伟达还为其首款中央处理器（CPU）Vera赢得了外部客户。SpaceX子公司SpaceXAI将部署该处理器。Vera采用英伟达设计的Olympus核心和Spatial Multithreading技术，完成任务的速度比x86 CPU快多达1.8倍。
第二季度边缘计算营收同比增长27%至72亿美元。英伟达还发布了与微软合作开发的RTX Spark，以及用于在Windows环境中开发和运行AI代理的DGX Station。它还推出了Alpamayo 2 Super，这是一个用于机器人出租车和自动驾驶汽车开发的开源推理模型。
英伟达预测第三季度营收中点值为1080亿美元，比华尔街预期的1041.9亿美元高出约3.7%。该展望不包括来自中国的数据中心营收。公司预计毛利率中点值为74%。
英伟达将向截至9月10日登记在册的股东支付每股0.25美元的第二季度现金股息。
尽管营收比去年同期增长了一倍多，英伟达股价在财报发布后下跌1.5%，收于209美元。该股随后在盘后交易中反弹至约218美元。
(Photo: Nvidia)
Nvidia posted record revenue for a 13th consecutive quarter as demand for artificial intelligence (AI) infrastructure continued to surge. Chief Executive Officer Jensen Huang said revenue could grow about 70% next year, underscoring his confidence in continued expansion of the AI market.
Nvidia said on Aug. 26 that revenue for its fiscal 2027 second quarter, ended July 26, rose 106% from a year earlier to $96.22 billion. Adjusted earnings per share (EPS) were $2.22, while gross margin was 75%. Gross margin increased 2.2 percentage points from a year earlier and edged up 0.1 percentage point from the previous quarter, slightly exceeding the company's prior guidance of 74.9%.
The results also topped Wall Street expectations. Analysts had projected second-quarter revenue of $92.17 billion and adjusted EPS of $2.10. Actual revenue was about 4.4% above the estimate, while EPS beat expectations by about 5.7%.
The data center business drove the quarterly performance. Data center revenue rose 117% from a year earlier to $89 billion. Nvidia began mass production of its Vera Rubin NVL72 last month and is supplying the systems to CoreWeave, Google Cloud, Microsoft Azure and Oracle Cloud. Vera Rubin NVL72 delivers as much as 30 times more throughput per watt than the previous-generation Blackwell platform, while reducing cost per token to about one-35th.
"Compute is revenue," Huang said, adding that the AI infrastructure buildout is proceeding at full speed and that Vera Rubin was built to support this moment.
Nvidia has also secured an external customer for Vera, its first central processing unit (CPU). SpaceXAI, a SpaceX subsidiary, will deploy the processor. Vera uses Nvidia-designed Olympus cores and Spatial Multithreading technology and can complete tasks as much as 1.8 times faster than x86 CPUs.
Second-quarter edge computing revenue rose 27% from a year earlier to $7.2 billion. Nvidia also unveiled RTX Spark, developed in collaboration with Microsoft, as well as a DGX Station for developing and running AI agents in Windows environments. It also introduced Alpamayo 2 Super, an open reasoning model for robotaxi and autonomous-vehicle development.
Nvidia forecast fiscal third-quarter revenue of $108 billion at the midpoint, about 3.7% above Wall Street's estimate of $104.19 billion. The outlook excludes data center revenue from China. The company projected gross margin of 74% at the midpoint.
Nvidia will pay a second-quarter cash dividend of $0.25 per share to shareholders of record as of Sept. 10.
Despite revenue more than doubling from a year earlier, Nvidia shares fell 1.5% after the earnings release to close at $209. The stock later rebounded in after-hours trading to around $218.