英伟达下调利润率展望，内存价格飙升推高成本关键事件
英伟达因内存芯片价格上涨导致成本增加，下调了毛利率展望。公司还预计营收增长将落后于需求增长，原因是晶圆代工产能的供应链瓶颈。
原文深度解读
本文报道了英伟达（Nvidia）在2027财年第二季度财报电话会议中下调毛利率展望，原因是内存芯片价格上涨推高成本。公司还预计营收增长将落后于需求增长，受晶圆代工产能、数据中心电力和可用站点等供应链瓶颈影响。文章涉及英伟达、SK海力士、三星电子、美光科技、谷歌、AWS、Anthropic、OpenAI等主体，关键数字包括毛利率预期（Q3 74%，Q4 71%-72%）、营收增长预期（2028财年约70%）、需求增长预期（100%）、超大规模资本支出（今年近8000亿美元，明年1.3万亿美元）、AWS部署200万GPU、OpenAI部署12GW基础设施等。与电子元器件供应链的关系：内存价格、代工产能、AI芯片需求直接影响半导体供应链的供需和成本结构。
- 英伟达在2026年8月26日的2027财年第二季度财报电话会议上，预计第三季度（8月至10月）毛利率为74%，第四季度（11月至1月）降至71%-72%，而第二季度毛利率为75%。
- 内存芯片价格上涨幅度超出英伟达此前预期，且明年价格将进一步上涨；英伟达正与三星电子、SK海力士和美光科技讨论扩大产能，预计影响在2028财年显现，届时毛利率预计恢复至72%-73%。
- 谷歌、AWS、Anthropic和OpenAI等主要客户正在设计针对其AI工作负载优化的ASIC芯片；英伟达认为ASIC是特定云或服务的推理芯片，而其平台支持数据处理、训练和推理。
- 英伟达预计2028财年营收同比增长约70%，而实际需求更高，明年需求增长预计达到100%。
- 五大超大规模数据中心今年的资本支出预计接近8000亿美元，明年达1.3万亿美元；AWS将额外部署200万块英伟达GPU。
- OpenAI计划到2030年部署12GW的AI基础设施；英伟达回应循环融资担忧，称OpenAI将成为历史上最大的科技公司。
供应链影响
- 内存芯片价格上涨可能推高英伟达及其他AI芯片制造商的成本，进而影响整个AI服务器供应链的定价和利润率。
- 英伟达与三星、SK海力士、美光的产能讨论可能缓解内存供应紧张，但具体效果取决于产能扩张速度和市场供需平衡。
- 晶圆代工产能瓶颈可能限制AI芯片出货量，影响下游服务器和数据中心建设进度，取决于代工厂扩产计划。
- 大型客户自研ASIC芯片可能分流部分AI推理需求，影响英伟达GPU的市场份额，但取决于ASIC的部署规模和性能表现。
- 超大规模数据中心和OpenAI的大规模资本支出可能持续拉动GPU、内存、电源等元器件需求，但实际影响取决于投资落地时间和供应链执行能力。
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。
SK集团董事长崔泰源（左）和英伟达CEO黄仁勋。（照片：THE ELEC）
英伟达因内存芯片价格上涨导致成本增加，下调了毛利率展望。公司还预计营收增长将落后于需求增长，原因是晶圆代工产能、数据中心电力和可用站点的供应链瓶颈。
英伟达在8月26日的2027财年第二季度财报电话会议上表示，第三季度（8月至10月）毛利率预计为74%，第四季度（11月至1月）将降至71%至72%。第二季度毛利率为75%。
内存芯片价格上涨预计将拖累盈利能力。英伟达表示，内存价格上涨幅度已超出其此前预期，且明年价格将进一步走高。公司还在与三星电子、SK海力士和美光科技讨论扩大产能。英伟达预计影响将在2028财年显现，届时毛利率预计恢复至72%至73%。
包括谷歌在内的大客户定制芯片开发是另一个挑战。谷歌、亚马逊网络服务（AWS）、Anthropic和OpenAI正在设计针对其AI工作负载优化的专用集成电路（ASIC）。然而，英伟达表示，ASIC是特定云或服务专用的推理芯片，而其平台支持数据处理、训练和推理。英伟达还表示，预计2028财年营收同比增长约70%，而实际需求更高。明年的需求增长预计将达到100%。
五大超大规模数据中心今年的资本支出预计接近8000亿美元，明年达1.3万亿美元。AWS将额外部署200万块英伟达图形处理器（GPU）。
OpenAI还计划到2030年部署12吉瓦（GW）的AI基础设施，使用英伟达产品。针对循环融资的担忧，英伟达表示OpenAI将成为历史上最大的科技公司，暗示其认为投资回收没有问题。
AI代理的需求正在推动扩张。AI代理所需的计算资源是人类员工执行任务的100倍，因为它们全天候运行并处理比传统AI更多的数据。英伟达预计AI代理最终将以100比1的比例超过员工数量，推动计算需求的持续增长。NeoCloud AI基础设施容量预计也将从去年的3吉瓦增加一倍以上，达到今年的8吉瓦。
英伟达报告2027财年第二季度营收为962.2亿美元，同比增长超过一倍，调整后每股收益（EPS）为2.22美元。这些数字分别超出市场预期约4.4%和5.7%。毛利率为75%，高于去年同期，略高于上一季度提供的74.9%展望。
SK Group Chairman Chey Tae-won, left, and Nvidia CEO Jensen Huang. (Photo: THE ELEC)
Nvidia lowered its gross-margin outlook as rising memory-chip prices increase costs. The company also expects revenue growth to trail demand growth because of supply-chain bottlenecks in foundry capacity, data-center power and available sites.
Nvidia said during its fiscal 2027 second-quarter earnings call on Aug. 26 that gross margin is expected to be 74% in the third quarter from August through October and fall to 71% to 72% in the fourth quarter from November through January. Gross margin was 75% in the second quarter.
Higher memory-chip prices are expected to weigh on profitability. Nvidia said the magnitude of memory price increases has exceeded its previous expectations and prices are headed even higher next year. The company is also in discussions with Samsung Electronics, SK hynix and Micron Technology to expand capacity. Nvidia expects the impact to become visible in fiscal 2028, when gross margin is projected to recover to 72% to 73%.
Custom chip development by major customers including Google is another challenge. Google, Amazon Web Services (AWS), Anthropic and OpenAI are designing application-specific integrated circuits (ASICs) optimized for their AI workloads. Nvidia, however, said ASICs are inference chips specialized for particular clouds or services, while its platform supports data processing, training and inference. Nvidia also said it expects fiscal 2028 revenue to grow about 70% from the previous year, while actual demand is higher. Demand growth next year is expected to reach 100%.
Capital expenditure by the five largest hyperscalers is expected to reach nearly $800 billion this year and $1.3 trillion next year. AWS will deploy an additional 2 million Nvidia graphics processing units (GPUs).
OpenAI also plans to deploy 12 gigawatts (GW) of AI infrastructure using Nvidia products by 2030. Addressing concerns over circular financing, Nvidia said OpenAI will grow into the largest technology company in history, suggesting it sees no problem recovering its investments.
Demand for AI agents is driving the expansion. AI agents require as much as 100 times more computing resources than human workers performing tasks because they operate around the clock and process more data than conventional AI. Nvidia expects AI agents eventually to outnumber employees by 100 to one, driving continued growth in computing demand. NeoCloud AI infrastructure capacity is also expected to more than double to 8 GW this year from 3 GW last year.
Nvidia reported fiscal 2027 second-quarter revenue of $96.22 billion, more than double a year earlier, and adjusted earnings per share (EPS) of $2.22. The figures beat market expectations by about 4.4% and 5.7%, respectively. Gross margin was 75%, up from a year earlier and slightly above the 74.9% outlook provided in the previous quarter.