BIS Places AI Bubble Burst on Systemic Risks List, Compares Credit to 2008 Collapse

The Bank for International Settlements (BIS) has placed the surge in investment in artificial intelligence on the same list of threats to global financial stability as persistent inflation and fiscal stress. The Basel-based institution published its Annual Economic Report this Sunday, June 28, and was explicit about the adverse scenario: "Disappointment in returns could trigger a sudden withdrawal of financing and transform the capex boom into a prolonged investment downturn, with potential effects on financial conditions."

The document asserts that the five largest hyperscalers are expected to spend over $1 trillion on AI capex between 2025 and 2026, an amount that exceeds both the profits and free cash flow of these companies and pushes some of them into debt issuance. For the BIS, the speed of the market has already decoupled from the effective economic returns of the technology.

The Explicit Parallel with 2008

The section that will dominate risk readings in the coming days is in the credit section. For the BIS, "the repricing of risk this time, whether triggered by higher interest rates or an AI collapse, has the potential to be equally disruptive" to the credit segment affected by the Global Financial Crisis of 2008. This comparison is not casual: the report links the risk to the financing architecture that has developed around AI infrastructure, particularly the growing role of hedge funds, private credit vehicles, and other non-bank intermediaries as alternative funding sources.

BIS Bulletin 120, authored by Iñaki Aldasoro, Sebastian Doerr, and Daniel Rees, complements the technical analysis. The authors document the migration of AI expenses from operating cash to debt and show that a significant portion of recent issuance in the United States has come from companies directly exposed to the sector.

Circular Financing: The Structure that Concerns the BIS

The report dedicates unusual space to what it calls "circular financing." The mechanics are described as follows: chip manufacturers and hyperscalers take equity stakes in AI labs or neocloud providers, and these same companies commit to multi-year contracts to purchase chips or computing capacity. "The terms are often not disclosed, with the risk that the same asset is pledged multiple times," states the document.

The BIS avoids naming names, but the design accurately describes the round that saw Nvidia and Microsoft taking stakes in labs that became multi-year clients of the investors themselves, in addition to partnerships between hyperscalers and dedicated capacity providers. The highlighted risk is the same as that of structured vehicles pre-2008: the stacking of exposures where the same promise of cash flow supports multiple liabilities.

Why the Warning Hits Wall Street at a Delicate Moment

The alert comes to a market already in adjustment. The Nasdaq has fallen more than 6% since its June 2 peak, and the BIS report arrives following a Bloomberg analysis published the day before about a renewed wave of tech debt issuance to finance AI capex. For Wall Street, the practical reading is that regulators now have a reference document to question banks about their aggregate exposure to AI infrastructure, both in direct credit and in prime brokerage for hedge funds operating in the sector.

In Europe, where the BIS has more direct influence over supervisors, the concern translates into pressure for disclosure. National central banks have already begun to request granularity on credit lines associated with data centers and specialized equipment. The combination of heavy capex, non-bank debt, and opaque multi-year contracts creates precisely the profile that prudential supervisors detest: limited visibility with high sector concentration.

The BIS does not claim that AI is a bubble. The point is more subtle. The institution argues that, even in a scenario where the technology delivers long-term productivity gains, the financial transition to get there brings a window in which the financing structure can break before the returns materialize. This is the central thesis of Aldasoro and colleagues in Bulletin 120: the shift from funding via cash to funding via debt changes the geometry of risk, even if the underlying asset remains valuable.

For CIOs of banks and professional services firms, the BIS report serves as a calibration signal. It is not the time to unwind already signed capacity contracts, but it is time to review exit clauses, exposure to single counterparties, and dependence on neocloud providers with underdeveloped balance sheets. The cost of a proactive renegotiation now is incomparably lower than the cost of discovering, amidst a repricing, that your compute provider is leveraged three levels over the same contract sustaining the service.