Central Bankers Warn the Artificial Intelligence Bubble Could Trigger a Global Financial Crash

What the BIS actually said

The Bank for International Settlements released its 2026 Annual Economic Report on Sunday, and the language on artificial intelligence was unusually direct. The BIS — which serves as the central bank for the world's central banks — warned that the artificial intelligence bubble forming around data center spending, chipmaker financing, and private credit structures poses a genuine threat to global financial stability.

The core concern is not just overvalued stocks. The BIS identified a specific transmission mechanism: if the five largest hyperscalers — Amazon, Alphabet, Microsoft, Meta, and Oracle — slow or halt their aggressive capital spending, companies across the AI supply chain could be unable to service their debt. That chain includes data center builders, cloud providers, neocloud operators, and chipmakers that have extended financing to their own customers.

BIS General Manager Pablo Hernandez de Cos warned that a reversal of what he called 'AI exuberance' could carry 'large macroeconomic consequences.' The report listed an AI bust alongside inflation and fiscal stress as the three most alarming threats to global prosperity right now.

The hidden debt structure nobody is pricing in

The BIS's sharpest concern is not hyperscaler balance sheets — it is what is not on them. Hyperscalers issued more than $120 billion in bonds in 2025 and routed another $120 billion through special purpose vehicles and private credit structures to fund data centers. That is shadow borrowing: economically equivalent to debt, but largely invisible to investors reading a standard earnings report.

The report also flagged what it called circular financing. In a typical arrangement, a chipmaker like Nvidia takes an equity stake in an AI lab. That lab then commits to multi-year chip purchases from the same chipmaker. The data center is outsourced to a third party that leases it back to the hyperscaler under a long-term contract with embedded exit clauses. The BIS noted that the terms of such deals are 'typically poorly disclosed,' with risks of the same asset being pledged multiple times.

Private credit funds — shadow banks that lend to companies outside the traditional banking system — have also piled into AI data center financing. The BIS report noted that signs of stress are already visible: some of these funds are inundated with redemption requests and have been forced to block withdrawals. That is not a future scenario. It is happening now.

The dotcom comparison — and where it gets worse

The BIS drew explicit parallels between the current AI investment boom and three historical precedents: the dotcom crash of the early 2000s, the British railway mania of the 1840s, and the Roaring 20s before the Great Depression. The scale of AI spending and the expectations of large productivity payoffs match the pattern.

But one element makes this cycle more structurally dangerous than dotcom. When the dotcom bubble burst, the physical infrastructure — fiber optic cables, server farms, routing equipment — retained value. It was eventually repurposed. The European Systemic Risk Board noted in December 2025 that AI overinvestment may not leave behind comparable durable assets. Software models, compute contracts, and leased capacity can become worthless quickly if demand evaporates.

The IMF has also made the dotcom comparison. The Bank of England warned in December 2025 that share prices were the 'most stretched' since the 2008 financial crisis. The BIS warning lands on top of that growing chorus. None of these institutions are calling the end of AI. All of them are flagging that the financing structure of the boom is opaque enough that a reversal could spread faster than most people expect.

What US builders and operators should take from this

For American developers, founders, and operators, the practical implication is uncomfortable. If your business depends on cheap cloud compute, expanding GPU access, or capital markets that reward AI bets generously, you are already exposed to the same macro risk the BIS is describing. You are not betting only on your product — you are also betting that hyperscalers keep spending through disappointment.

The FSOC — the Financial Stability Oversight Council — launched an interagency AI working group in December 2025 to monitor exactly these risks. Senate Democrats wrote to the FSOC in January 2026 urging a formal investigation into AI-related debt growth. Former Treasury Secretary Janet Yellen flagged concerns about circular financing and hype-driven AI valuations. JPMorgan CEO Jamie Dimon has referenced 'inflated AI assets' and warned that 'people are doing dumb things' amid the bubble fears.

The market is already giving signals. The Nasdaq dropped as much as 1.3% last Friday. Tech stocks tied to the AI boom have swung sharply. South Korea's chipmaker-heavy index saw swings exceeding 10% in a single day. SpaceX's IPO shares fell 25% from their peak within days. Google has reportedly capped Meta's access to Gemini due to capacity constraints — a sign that even at the frontier, supply is not unlimited.

None of this means the AI boom is over. The BIS itself acknowledged that AI spending supported global growth in 2025. But the report's core warning is structural: the financial system is now tightly coupled to AI capex decisions made by five companies. If those decisions change, the feedback loop is fast, deep, and poorly visible to regulators and investors alike.

Sources