AI spending surge faces financial strain, watchdog warns

The Bank for International Settlements warned that the AI investment boom could turn into a broader financial problem if expected returns fall short. In its Annual Economic Report 2026, the Basel-based institution said the scale of spending by major cloud and technology companies resembles past investment surges that ended in recessions.

The BIS said the five largest hyperscalers are on track to spend more than $1 trillion on AI-related capital expenditures across 2025 and 2026. According to the report, that pace has begun to exceed their earnings and free cash flow, leading some companies to use debt to fill the gap.

Big spending, uncertain returns

The BIS said AI has shown measurable productivity gains in task-level studies, including time savings of 20% to 50%. Its concern is that several dominant firms are making large, overlapping bets at the same time because they expect only a small number of winners to control the market.

Using contest-theory modeling, BIS economists said rising competitive pressure can push companies to spend more while reducing the net gains for the sector after investment costs. In weaker scenarios, the report said, those gains could turn negative and cause lenders or investors to pull back.

The BIS compared the AI buildout with earlier technology-driven booms, including canal investment in the 1830s, British railways in the 1840s and the dot-com cycle that ended in 2000. The report said those episodes began with real advances but produced investment reversals when commercial returns failed to match expectations.

Financing risks spread beyond tech

The BIS said the financing behind AI infrastructure adds to the risk. It described a network of private deals linking hyperscalers, chipmakers, AI labs, data center contractors and lenders, including arrangements in which companies invest in AI labs that then agree to buy chips or computing services.

The report said many of those deal terms are not well disclosed and warned that the same assets may be pledged more than once. It also said third-party data center contractors often lease facilities back to customers under long contracts that may contain exit provisions.

If hyperscalers reduce spending, the BIS said companies across the supply chain could lose revenue at the same time. The report identified engineering and construction firms as especially exposed because it said their balance sheets are comparatively weak.

Zhang Tao, the BIS representative for Asia and the Pacific, told the South China Morning Post that a correction could unfold faster than earlier banking crises because much of the funding runs through hedge funds and private credit vehicles, which face less oversight than banks.

Apollo Global Management Chief Economist Torsten Slok argued in mid-May that AI had spread through financial markets beyond equities. According to Apollo figures cited in the report, AI is tied to nearly half of investment-grade bond issuance, 87% of venture capital funding and an increasing share of high-yield debt.

Market fallout and policy pressure

The BIS said a sharp fall in AI-linked stocks could hit consumer spending through household wealth. The report said U.S. stocks make up about 64% of the MSCI Global index, while household equity exposure relative to income has more than doubled since 2010.

The report also pointed to an energy shock after the Strait of Hormuz closed following the start of the Iran conflict in late February. According to the BIS, the disruption removed more than 10 million barrels of crude oil per day from global supply, sent oil as high as $120 a barrel within two weeks and lifted global headline inflation by half a percentage point.

The BIS said policymakers should focus on financial robustness by keeping watch on inflation, rebuilding government fiscal capacity and applying prudential standards to non-bank lenders involved in AI financing.

This story draws on original reporting from Fortune.