美国科技巨头为支撑人工智能基础设施建设而进行的大规模融资正在加速，但市场已显现疲劳迹象。根据研究数据，这些企业的隐性债务在过去四年内增长了八倍，达到1.65万亿美元，远超其账面披露的1.35万亿美元债务规模[1]。其中，美国非财务报表债务约为1.2万亿美元，超过820亿美元来自在建数据中心[1]。
债券发行步伐创下纪录。根据标普全球统计，超大规模企业及相关实体在2026年上半年发行了225亿美元债券，同比增幅达973.7%，全年发行规模预计达到400亿美元[1]。然而，经济学家和分析机构对这一趋势的可持续性提出了警告。美国RSM公司首席经济学家约瑟夫·布鲁苏埃拉斯表示，这种情况不会无限期地持续下去[1]。Capital Economics的分析进一步指出，如果债务趋势保持不变，企业和政府债券发行占GDP的比例将创下除疫情年份外有记录以来的最高水平[1]。与此同时，联邦政府2026财年的预算赤字预计接近2万亿美元[1]，这意味着政府和私人部门的债务将同时激增，投资者最终的融资意愿恐难以维系。
Mega-cap artificial intelligence companies are on a record debt-issuance pace that experts warn cannot continue indefinitely. In the first half of 2026 alone, hyperscale enterprises and related entities issued $22.5 billion in bonds, a 973.7% increase compared to the same period in the previous year, with full-year issuance projected to reach $40 billion [1]. This aggressive borrowing spree masks a far larger hidden debt problem: according to research cited, U.S. technology giants have accumulated $1.65 trillion in off-balance-sheet debt over four years—a roughly eightfold expansion that now exceeds their reported on-the-books debt of $1.35 trillion [1].
The scale of undisclosed obligations reflects the financing mechanisms driving infrastructure buildout. Moody's estimates place non-financial statement debt at approximately $1.2 trillion, with over $82 billion coming from data centers still under construction [1]. As these debt levels surge in tandem with government obligations, market conditions are tightening. The U.S. federal government alone faces a projected budget deficit approaching $2 trillion for fiscal year 2026 [1]. Capital Economics analysis suggests that if current debt issuance trends persist, combined corporate and government bond issuance will reach historic highs relative to GDP, excluding pandemic years [1].
RSM's chief economist Joseph Brusuelas captured the underlying constraint plainly: "That will not endure indefinitely" [1]. With both private and public sectors competing for capital simultaneously, investor appetite for further leverage will eventually reach a breaking point, forcing a reassessment of how AI infrastructure expansion is financed.