美光再创纪录：内存短缺“看不到尽头”关键事件
美光科技（Micron Technology）季度财报再次刷新历史纪录，受人工智能（AI）推动的内存需求持续扩大所带动。原文称内存需求仍集中于相关领域，并援引“内存短缺看不到尽头”的说法。
原文深度解读
本文报道美光科技（Micron Technology）在2026财年第四季度（6-8月）再次刷新季度业绩纪录，营收和营业利润分别同比增长约4.8倍和12倍，主要受人工智能（AI）数据中心内存需求推动。文章涉及美光、其CEO Sanjay Mehrotra、云内存业务部门、核心数据中心业务部门，以及HBM、DRAM、战略客户协议（SCA）等关键要素，并提及美国爱达荷州ID1、ID2晶圆厂和纽约新厂的产能扩张计划。与电子元器件供应链的关系在于：文章直接描述了内存（尤其是HBM和DRAM）供需持续紧张、长期供应协议扩大、产能建设周期等影响上游存储器件供应格局的信息。
- 美光2026财年第四季度营收为542.29亿美元，同比增长379%（去年同期为113.15亿美元），营业利润为437.51亿美元，约为去年同期的12倍（去年同期为36.54亿美元），这是自2025年第三季度以来连续第六个季度创纪录。
- 营收远超LSEG汇编的市场预期510.7亿美元；云内存业务部门和核心数据中心业务部门营收分别为163亿美元和180亿美元，两者合计占总营收60%以上。
- 美光给出2027财年第一季度营收指引为600亿至630亿美元，超过市场预期的约570亿美元；CEO表示2027和2028财年内存与存储供需将比2026年紧张得多，且看不到供需恢复平衡的时间。
- HBM需求被视为加剧供应短缺的主要因素：预计到2028年行业HBM位元出货量增速将快于传统DRAM，且明年大部分HBM产能已与客户签约，合同价格同比大幅上涨。
- 长期供应协议方面，美光已签署26份战略客户协议（SCA），客户承诺规模达320亿美元，较第三季度财报时披露的16份协议、220亿美元显著扩大。
- 产能扩张计划包括：美国爱达荷州ID1晶圆厂计划2027年中开始晶圆生产，ID2晶圆厂计划2028年底开始；纽约新厂近期进入破土动工阶段，目标2030年初首次晶圆生产。公司还预计未来AI内存需求将从数据中心扩展至自动驾驶汽车和人形机器人等‘物理AI’领域。
供应链影响
- 由于美光表示2027和2028财年内存与存储供需可能比2026年更紧张，且看不到恢复平衡的时间，存储器件（尤其是HBM和DRAM）的供应紧张态势可能延续至2028年。
- HBM明年大部分产能已与客户签约且合同价格同比大幅上涨，可能意味着非签约客户或中小型买家获取HBM供应的难度和成本可能上升，具体取决于后续产能释放节奏。
- 战略客户协议数量和承诺规模从16份/220亿美元增至26份/320亿美元，可能表明下游客户正在通过长期协议锁定供应，供应链中的长期合约占比可能提高。
- 爱达荷州ID1、ID2及纽约新厂的投产时间分别指向2027年中、2028年底和2030年初，新增晶圆产能对缓解供应紧张的实际效果可能取决于这些工厂的建设进度和爬坡速度。
- 公司预计AI内存需求可能从数据中心扩展至自动驾驶汽车和人形机器人等‘物理AI’领域，若该趋势实现，可能进一步扩大存储器件在非数据中心应用中的需求来源。
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。
美光在美国爱达荷州正在建设的新内存工厂（照片=美光YouTube截图）
美光科技（Micron Technology）再次打破其历史季度盈利纪录，受人工智能（AI）推动的内存需求持续扩大所带动。随着内存需求继续集中于AI数据中心，该公司预计业绩增长将持续到明年。
根据美光于30日（当地时间）披露的2026财年第四季度（6-8月）业绩，营收和营业利润分别同比增长约4.8倍和12倍。
营收达到542.29亿美元，较去年同期（113.15亿美元）增长379%。这标志着自2025年第三季度以来连续第六个季度创下纪录。它也远超伦敦证券交易所集团（LSEG）汇编的市场预期510.7亿美元。
同期，营业利润飙升至437.51亿美元，约为去年（36.54亿美元）的12倍。
推动盈利增长的无疑是AI数据中心的内存需求。包含高带宽内存（HBM）销售的云内存业务部门和核心数据中心业务部门营收分别为163亿美元和180亿美元。仅这两个业务部门就占总营收的60%以上。
该公司还对其明年的财务前景表示信心。美光给出的2027财年第一季度营收指引为600亿至630亿美元。这超过了市场预期（约570亿美元）。
美光CEO Sanjay Mehrotra在财报电话会议上预测：“自上次财报以来，行业需求进一步走强，2027财年和2028财年的内存与存储供需形势将比2026年紧张得多。”
他补充说：“尽管需求趋势强劲——包括客户提出的新的上调请求——并且计划增加DRAM洁净室空间，但仍看不到供需何时恢复平衡。”
尤其是HBM需求被列为加剧供应短缺的主要因素。Mehrotra提出的证据是，预计到2028年行业HBM位元出货量将快于传统DRAM增长，以及明年大部分HBM产能已与客户签约。他解释说，合同价格也同比大幅上涨。
通过长期供应协议，盈利可见度也在上升。迄今为止，该公司已签署26份战略客户协议（SCA），客户承诺规模达到320亿美元。这较第三季度财报时报告的16份协议和220亿美元显著扩大。
为应对中长期需求，产能也在扩张。美国爱达荷州的ID1晶圆厂计划于2027年中开始晶圆生产，ID2晶圆厂计划于2028年底开始。纽约新厂近期进入破土动工阶段，目标是在2030年初首次晶圆生产。
该公司预计，未来AI内存需求将从数据中心扩展至自动驾驶汽车和人形机器人等‘物理AI’领域。
“日益复杂的系统需要更高性能和更节能的内存与存储来实时运行，”Mehrotra说。“人形机器人预计也需要与自动驾驶汽车类似水平的内存和存储。”
美光季度业绩趋势。（来源=美光，基于美国GAAP，单位：美元）
A new memory plant under construction by Micron in Idaho, U.S. (Photo = Micron YouTube capture)
Micron Technology has broken its all-time quarterly earnings record once again, driven by expanding demand for memory fueled by artificial intelligence (AI). With memory demand continuing centered on AI data centers, the company projected that performance growth will persist into next year.
According to fiscal fourth-quarter 2026 (June-August) results disclosed by Micron on the 30th (local time), revenue and operating profit surged approximately 4.8 times and 12 times, respectively, year-over-year.
Revenue reached $54.229 billion, up 379% compared to the same period last year ($11.315 billion). This marks the sixth consecutive quarter of record-breaking performance since the third quarter of 2025. It also far surpassed the market forecast of $51.07 billion compiled by the London Stock Exchange Group (LSEG).
During the same period, operating profit soared to $43.751 billion, about 12 times higher than the previous year ($3.654 billion).
What drove the earnings growth was unquestionably memory demand for AI data centers. Revenue for the Cloud Memory business unit, which includes high bandwidth memory (HBM) sales, and the Core Data Center business unit recorded $16.3 billion and $18 billion, respectively. These two business units alone accounted for more than 60% of total revenue.
The company also expressed confidence in its financial outlook for next year. Micron presented its revenue guidance for the first quarter of fiscal 2027 at $60 billion to $63 billion. This exceeds the market expectation (approximately $57 billion).
Sanjay Mehrotra, CEO at Micron, predicted during the earnings call that “industry demand has strengthened further since the last earnings report, and the memory and storage supply and demand situation in fiscal 2027 and fiscal 2028 will be much tighter than in 2026.”
He added, “Despite robust demand trends—including new upside requests from customers—and plans for additional DRAM cleanroom space, there is no visibility into when supply and demand will regain balance.”
In particular, HBM demand was cited as a major factor exacerbating the supply shortage. Mehrotra presented as evidence the expectation that industry HBM bit shipments will grow faster than traditional DRAM through 2028, along with the fact that most of next year's HBM volume is already contracted with customers. Contract prices have also risen significantly year-over-year, he explained.
Earnings visibility is also rising through long-term supply agreements. To date, the company has signed 26 Strategic Customer Agreements (SCAs), with customer commitment scale reaching $32 billion. This is a significant expansion from the 16 agreements and $22 billion reported during the third-quarter earnings announcement.
To respond to medium- to long-term demand, production capacity is also being expanded. The ID1 fab in Idaho, U.S., is scheduled to begin wafer production in mid-2027, and the ID2 fab in late 2028. The new fab in New York recently entered the groundbreaking phase, targeting initial wafer production in early 2030.
The company expected that future demand for AI memory will expand beyond data centers to 'Physical AI,' such as autonomous vehicles and humanoid robots.
“Increasingly complex systems require much higher performance and more power-efficient memory and storage to operate in real time,” Mehrotra said. “Humanoid robots are also expected to require levels of memory and storage similar to those of autonomous vehicles.”
Micron's quarterly performance trend. (Source = Micron, based on U.S. GAAP, Unit: USD)