摩根士丹利上调预测：2027年全球AI资本开支将达1.58万亿美元 英伟达独占约6900亿美元
摩根士丹利10月18日发布报告，将2027年全球 AI 相关资本开支预测上调14%，至约1.58万亿美元（约1.6万亿）。此次上调主要因为新计入了此前未统计的数据中心租赁承诺。
报告预计，英伟达将成为最大赢家，独占约6,900亿美元——相当于每10美元 AI 基建投入中，就有4美元以上流向英伟达的数据中心硬件。标普500成分公司将贡献其中约1.03万亿美元（占65%）；数据中心租赁承诺在2027年将达到1,950亿美元，其中甲骨文一家的租赁承诺达572亿美元，其次为微软、Meta、亚马逊和 Alphabet。
值得注意的是，这份乐观预测发布之际，AI 信用市场正在亮起黄灯：为 AI 基建举债的多家公司债券已跌入困境区间，摩根士丹利同时警告，超大规模云厂商与新兴算力公司2027年可能烧掉约6,000亿美元自由现金流。乐观的资本开支预测与吃紧的融资环境形成对照。（GSM 据摩根士丹利报告及多家财经媒体报道）
Morgan Stanley Raises 2027 Global AI Capex Forecast to $1.58 Trillion; Nvidia Seen Capturing About $690 Billion
Morgan Stanley raised its forecast for 2027 global AI-related capital expenditure by 14 percent to about $1.58 trillion in a report released October 18, after folding previously uncounted data-center lease obligations into its totals.
The bank expects Nvidia to be the biggest winner, capturing roughly $690 billion — more than four of every ten AI infrastructure dollars flowing to Nvidia's data center hardware. S&P 500 companies are projected to account for about $1.03 trillion of the spending, or 65 percent. Data center lease commitments are estimated at $195 billion in 2027, led by Oracle at $57.2 billion, followed by Microsoft, Meta, Amazon and Alphabet.
The upbeat forecast lands as credit markets flash warnings: bonds of several companies borrowing heavily for AI infrastructure have fallen into distressed territory, and Morgan Stanley warns hyperscalers and neoclouds could burn about $600 billion in free cash flow in 2027. (GSM, based on Morgan Stanley's report and financial media coverage)
来源：GSM·美国（原文链接）
图片来源：图：证券交易所行情中心（资料图）。图片来源：Wikimedia Commons，CC BY 2.0。Photo: Stock exchange market centre (file photo). Source: Wikimedia Commons, CC BY 2.0.
