摩根士丹利称，超大规模云厂商的资本开支并未放缓

Morgan Stanley states that capital expenditure by hyperscale cloud vendors has not slowed

【环球网财经综合报道】日前，摩根士丹利最新发布的全球云资本支出追踪报告显示，全球最大超大规模云厂商的资本开支并未放缓，预计2027年同比增幅将达29%。

[Global Network Financial Comprehensive Report] A few days ago, Morgan Stanley released its latest Global Cloud Capital Expenditure Tracking Report. It shows that capital expenditure by the world’s leading hyperscale cloud vendors has maintained growth, with the year-on-year growth rate projected to hit 29% in 2027.

摩根士丹利指出，市场对2027年云资本开支的一致预期为1.2万亿美元，但该预测可能偏保守，实际支出规模或攀升至1.4万亿美元。过去一年，投资者对AI数据中心大规模建设所伴随的资本开支飙升持续担忧，给市场带来一定波动。

Morgan Stanley notes that the market consensus for cloud capital expenditure in 2027 stands at $1.2 trillion, but this forecast may be overly conservative, with actual spending potentially rising to $1.4 trillion. Over the past year, investors remain concerned about the soaring capital expenditure associated with the large-scale construction of AI data centers, which has brought some volatility to the market.

然而，算力需求始终超出供给，AI商业化变现能力也在持续改善。在此背景下，全球四大超大规模云厂商中有三家已在2026财年上调资本开支指引：亚马逊从2000亿美元上调至2200亿美元；谷歌将指引区间从1800亿-1900亿美元收窄至1950亿-2050亿美元；Meta从1250亿-1450亿美元区间收窄至1300亿-1450亿美元。微软则维持1900亿美元不变。

However, demand for computing power has consistently outstripped supply, and the commercialization capacity of AI keeps improving. Against this backdrop, three of the world’s four leading hyperscale cloud providers have raised their capital expenditure guidance for fiscal year 2026: Amazon lifted its target from $2000 billion to $2200 billion. Google narrowed its guidance range from $1800–$1900 billion to $1850–$2050 billion. Meta tightened its range from $1250–$1450 billion to $1300–$1450 billion. Microsoft, meanwhile, maintained its guidance at $1900 billion.

爱建证券也发布研报认为，海外头部云厂商持续布局AI算力基础设施，中长期有望持续拉动GPU、高速存储等上游硬件需求，拉动HBM、服务器DRAM需求增长，海外存储大厂产能向AI高端存储倾斜。DRAM/NAND价格上涨节奏边际放缓，行业结构性分化凸显。

AllianceBernstein also released a research report, stating that overseas leading cloud vendors continue to build out AI computing infrastructure. In the medium and long run, this is expected to keep driving demand for upstream hardware such as GPUs and high-speed storage, and boost demand for HBM and server DRAM. Overseas major storage manufacturers are shifting their production capacity toward high-end AI storage. The marginal growth rate of DRAM and NAND price hikes has moderated, highlighting structural differentiation within the industry.