AI 风险暴露已集中于大型上市股票，且正愈来愈多地延伸到信贷：Alphabet、Amazon、Meta、Microsoft、Nvidia 与 Oracle 合计约占标普 500 指数近 $60 trillion 市值的约 25%，而且它们也正进入债务市场，为巨额资本支出筹资；文中以 Meta 的路易斯安那资料中心融资作为例子，称其为一笔 $30 billion 的交易，涉及单笔规模最大的公司债发行。同时，估值风险以 Nvidia 作为示例：若其本益比倍数仅回落至标普 500 的平均水准，将意味著市值约下滑 $1.5 trillion；文章指出，危机严重程度与其说取决于损失本身，不如说取决于杠杆及其所在位置，并对比 2000 年代初的网路泡沫破裂与次贷危机，还提到 2021 年的 Archegos，当杠杆部位崩解时造成放贷方损失超过 $10 billion。
该社论的主要担忧是，AI 融资分散在多种工具之中，可能掩盖最终由谁承担风险，以及曝险如何快速转移：私募信贷机构估计已放贷 $200 billion（其中部分资金可能向银行借来），另有数百亿被打包成具分层（tranche）结构的证券化产品，而一些参与私下交易的保险公司则被描述为愈来愈依赖短期融资。外溢效应可能来自对易受 AI 颠覆的软体公司所背负的「hundreds of billions」债务，亦可能透过非信贷渠道出现，例如劳动市场的扰动与 AI 驱动的交易失误；作为政策回应，它呼吁搜集资料以辨识杠杆集中处，并以更高的股权资本作为韧性缓冲，同时警告监管者近来已放松股权要求、移除杠杆贷款的限制，并淡化对系统性监测的重视。
The AI boom is being financed at unprecedented scale in US capital markets, with “hyperscalers” such as Meta Platforms Inc. expected to invest more than $3 trillion through 2030 in data and power infrastructure, an effort portrayed as vastly larger than the Manhattan Project but funded by private shareholders and creditors; the editorial argues regulators should ask now how the system and broader economy would cope if a boom-to-bust cycle hits while there is still time to adapt.
AI exposure is already concentrated in major public equities and increasingly in credit: Alphabet, Amazon, Meta, Microsoft, Nvidia, and Oracle together are about 25% of the S&P 500 Index’s nearly $60 trillion market capitalization, and they are also pushing into debt markets to fund large capital expenditures; Meta’s Louisiana data center financing is cited as a $30 billion deal involving the largest single corporate bond issuance, while valuation risk is illustrated by Nvidia, where a price-to-earnings multiple falling merely to the S&P 500 average would imply about a $1.5 trillion drop in market capitalization; crisis severity, the piece notes, depends less on losses per se and more on leverage and where it sits, contrasting the early-2000s dot-com bust with the subprime bust and pointing to Archegos in 2021, which generated more than $10 billion in lender losses when leveraged positions unraveled.
The editorial’s main concern is that AI financing is fragmented across instruments that can obscure who ultimately bears risk and how quickly exposures can shift: private-credit firms have loaned an estimated $200 billion (with some funding likely borrowed from banks), tens of billions more are packaged into securitizations with tranche structures, and some insurers involved in private deals are described as increasingly reliant on short-term financing; spillovers could come from “hundreds of billions” in debt on software companies vulnerable to AI disruption, plus non-credit channels such as labor-market disruption and AI-driven trading failures; as a policy response, it calls for data-gathering to identify leverage concentrations and for higher equity capital as a resilience buffer, warning that regulators have recently relaxed equity requirements, removed leveraged-lending curbs, and de-emphasized systemic monitoring.