AI 的债务狂欢难以为继,隐性借款已达 1.65 万亿美元。
AI's debt binge can't last, hidden borrowing reaches $1.65T

原始链接: https://fortune.com/2026/07/31/ai-debt-hypescalers-capex-capital-spending-hidden-borrowing-bond-issuance/

科技巨头们正通过大规模借贷掀起一场前所未有的AI投资狂潮,今年迄今已发行了2250亿美元的债券。尽管投资者基本消化了这一激增,但随着企业融资成本的提高,市场疲劳感正逐渐显现。这场“债务狂欢”因美国政府自身巨大的借贷需求而变得更加复杂,给私营部门的流动性带来了额外压力。 除了公开的债券发行,科技巨头还在积累巨额的“隐性债务”——估计在1.2万亿至1.65万亿美元之间——这些债务源于长期数据中心租赁和设备采购协议等资产负债表外承诺。虽然这些公司仍保持着强劲的信用评级,但其商业模式正从“轻资产”软件架构转变为资本密集型基础设施运营。专家警告称,这种对大规模杠杆的依赖不太可能长期持续;随着联邦赤字和企业借贷持续攀升,资本流向最终可能不再那么充裕,这将迫使人们重新审视AI繁荣背后的真实成本。

这篇 Hacker News 帖子讨论了《财富》杂志关于支撑人工智能产业的 1.65 万亿美元“隐性”债务的报道。 评论者对该模式的可持续性看法严重分歧。怀疑论者将其与互联网泡沫进行对比,警告称尽管技术可能具有变革性,但当前的金融格局——以巨额资本支出和过度企业借贷为特征——已变得岌岌可危。他们担心如果投资者热情减退,可能会出现“抵押贷款支持证券(MBS)式”的崩溃,并指出美联储已不再像过去那样吸收债务。 相反,一些人认为人工智能应用仍处于早期阶段,当前的计算资源限制证明了这些支出的合理性。支持者认为,即使发生类似互联网时代的“炒作周期”崩盘,也不会标志着该技术的终结,而是一次必要的市场调整。许多参与者对企业领导层持怀疑态度,认为高管们将短期关键绩效指标(KPI)和奖金置于长期财务健康之上,一旦泡沫破裂,公众可能会暴露在“大而不能倒”的风险之下。
相关文章

原文

AI’s insatiable need for debt has so far been matched by investors’ appetite for it, but they may turn nauseous on the belly-busting volumes coming from tech giants.

The latest quarterly reports from AI hyperscalers show that their massive spending plans remain on track, with Amazon even raising its capital expenditure guidance. That means even more bond issuance is on the way after an already staggering debt orgy.

The numbers paint a picture of a borrowing binge that’s bigger—and murkier—than it looks on paper. S&P Global counts $225 billion in bonds issued by hyperscalers and related entities like Nvidia so far this year, putting them on pace for a record haul in 2026—but that’s just the visible debt. Other analyses suggest so-called hidden debt at the five U.S. tech giants has ballooned, meaning the AI boom’s true price tag is only partly reflected in the bond markets that everyone’s watching.

Here is what is visible—and just barely visible—in the hyperscalers’ debt loads.

‘Market participants are growing leery’

S&P Global calculated that hyperscalers and “related entities” like Nvidia have issued $225 billion in bonds so far in 2026, representing a 973.7% jump through midyear. They are on pace to issue $400 billion for the full year.

But markets are showing signs of fatigue, after absorbing this flood of debt in such a short time, S&P warned, pointing out that hyperscalers are now paying a higher premium compared with yields on risk-free bonds.

“Market participants are growing leery of quickly rising leverage from issuers previously characterized by strong and reliable cash flow,” the report said.

At the same time, the federal government also needs bond investors to scoop up all the debt coming out of the Treasury Department, with the budget deficit this fiscal year expected to hit nearly $2 trillion.

And unlike earlier periods of heavy debt, the Federal Reserve is no longer a big buyer of Treasuries, placing a heavy burden on private-sector investors. 

S&P Global

Capital Economics noted that if debt trends from the first half of this year continue into the second half, then total corporate and government bond issuance as a share of GDP will be more than any year on record outside the pandemic.

RSM chief economist Joseph Brusuelas said in a note last week that demand for both types of debt remains strong for now.

“Yet that will not endure indefinitely,” he added.

Federal deficits will eventually cause lenders to charge a higher premium on public and private borrowers, Brusuelas predicted.

To be sure, those yields will still attract investors looking for bigger returns, but he also cautioned against complacency.

“At some point, the rivers of capital financing private and government debt issuance will flow less freely,” Brusuelas wrote.

Meanwhile, the official bond tally that’s hitting the market understates all the actual borrowing that’s going on to fund the AI boom.

Explosion in hidden debt

According to a study by Nikkei, so-called hidden debt at U.S. tech giants has exploded by 8x in just four years to $1.65 trillion. That amount doesn’t appear on balance sheets and even exceeds the $1.35 trillion in debt that does appear on their books.

These hidden debts can consist of tech companies signing long-term purchase deals for graphics processing units and servers, or lease agreements with data center operators, the report said.

They are legitimate practices under accounting rules and are not totally hidden as they are often disclosed in annotations in quarterly financial statements, rather than on the balance sheet. Much of the hidden debt will also become official at some point, especially when data centers start operations.

Similarly, Moody’s flagged off-balance-sheet deals in a recent report, putting them at $1.2 trillion, with more than $820 billion of that attributed to data centers that are still under construction.

The credit rating agency described them as debt-equivalent liabilities that will leave companies on the hook for significant rent payments in the future.

Despite all the obligations, Moody’s said hyperscalers still have some of the most robust balance sheets in the corporate world, and their investment-grade ratings are not facing imminent risk.

Still, the tech giants are undergoing a fundamental shift, as seen by their relentless spending and borrowing.

“Previously, these companies relied on asset-light structures centered on software, intellectual property, and scalable cloud services that required modest capital investment,” Moody’s said. “The transition from asset-light to asset-heavy models requires unprecedented levels of investment and capital raising.”

联系我们 contact @ memedata.com