The ECB detects parallels between debt tied to AI and the dot-com and subprime bubbles

The ECB warns of the growing opaque debt that finances AI and compares it to the excesses of the dot-com bubbles and subprime mortgages.

3 minutes

fotonoticia 20261118144929 1920
Add DEMÓCRATA to Google

Published

3 minutes

The unprecedented scale of the investments planned for the coming years across the entire supply chain of artificial intelligence is pushing many companies toward increasingly complex and off-balance-sheet financing formulas, according to the European Central Bank (ECB), which sees “historical parallels” with the dot-com bubbles and the subprime mortgages.

In its latest “Financial Stability Report,” the ECB’s experts point out that the cumulative capital expenditure of the major U.S. tech companies could amount to around $5.5 trillion (€5 trillion) by 2030. This would imply annual AI investment needs of more than 3% of U.S. GDP between 2025 and 2032, “a figure considerably higher than the financing of other major innovations in the past.”

The document emphasizes that this investment push in AI is concentrated above all in a small core of U.S. companies, including the large hyperscalers and semiconductor manufacturers. According to the ECB, these firms began recording negative free cash flows in 2026, which has led them to rely to a greater extent on external financing, both through capital increases and through more debt.

In addition to turning to public and private debt markets, the ECB warns that “companies are increasingly using opaque off-balance-sheet structures” to fund the deployment of AI infrastructure. These formulas complement bond issuance with other mechanisms supported by expected future cash flows to be obtained from monetizing that infrastructure.

In practice, the big tech companies sign contractual agreements to use, for years, the AI infrastructure of data center developers and owners of advanced chips. The future cash flows associated with those contracts serve as collateral for owners of AI-linked assets to seek financing from banks, bond buyers, or private credit funds.

"These cash flow commitments are similar to debt in economic terms, but they do not appear as liabilities in the balance sheets of large technology companies under U.S. accounting standards," warns the ECB, which calculates that these companies had accumulated off-balance-sheet commitments of more than $4 trillion (€3.6 trillion) in the second quarter of 2026, an amount far higher than their recorded on-balance-sheet liabilities.

According to the institution, "the opacity of AI-related debt is exacerbated by circular links in the form of mutual commitments and interconnected holdings among the companies in the AI supply chain." In this regard, it recalls that many chip purchases by AI companies are financed directly through the microprocessor manufacturers themselves, which contributes to inflating the revenue forecasts of all the parties involved.

In turn, the hyperscalers maintain connections with other companies in the AI ecosystem through loans and equity stakes. The capital gains derived from the revaluation of these private equity investments, especially in AI laboratories, represent an increasing share of the profits that major cloud service providers report.

"The rapid shift toward debt financing evokes historical parallels with the dot-com bubbles and subprime mortgages, and could generate contagion effects that affect the financial stability of the euro area," economists at the ECB warn. In their view, even in the scenario in which AI manages to fundamentally transform the global economy, its leveraged expansion could lead to excess capacity, inefficient capital allocation, and credit losses for those providing the financing.

At the same time, the ECB report finds that investors are starting to show greater sensitivity to the risks associated with the volume and pace of debt growth needed to drive AI, as well as to the lack of transparency and the circular nature of many of these financing structures.