
The BIS warns Big Tech's $1 trillion AI spending binge risks ending in a bust that could rattle markets and damage the global economy. Margin debt and circular financing amplify the danger.
The Bank for International Settlements warned in its annual economic report that a sharp pullback in Big Tech’s AI investment could trigger a prolonged investment drought, economic contraction, and even a financial crisis. The BIS, the central bank for central banks, has a track record of flagging systemic risks before they hit. It was the BIS that warned about dangerously elevated housing prices years before 2008. Alan Greenspan dismissed those concerns then. This time the target is the $1 trillion spending spree by the five largest hyperscalers between 2025 and the end of 2026.
The warning comes as corporate credit spreads hover near their lowest levels in two decades, making debt cheap for tech issuers. The Financial Times, which made the BIS report its lead story, noted that tech groups have flooded the global credit market with hundreds of billions of dollars in bonds. Allianz’s chief investment officer described SpaceX’s decision to launch a $25 billion bond sale so soon after its IPO as proof of “bubble territory.” The BIS report itself is more measured but no less stark: worse-than-expected returns in the AI sector could prompt investors to rapidly curb financing, and the resulting equity market correction could hit broader wealth because households now have a larger share of equities relative to their income and net worth than in any past AI cycle.
The BIS points to two specific financial vulnerabilities. The first is circular financing. One AI company’s capital spending becomes another’s revenue, which funds more spending in a closed loop. That structure mirrors the trust-of-trusts leverage of the 1920s and the collateralized debt obligations that amplified the 2008 crisis. The second is margin debt. The current level of margin debt relative to market capitalization sits above the dot-com peak and is not far from 1929 readings, according to FINRA data cited in the report. A correction that triggers forced selling would hit the hyperscalers that dominate index weightings and then cascade into the broader market. The BIS estimates that a 30% Nasdaq drop would reduce U.S. household net worth by roughly $4 trillion.
How the AI Investment Machine Could Unwind
The transmission path the BIS outlines runs through debt markets. Hyperscalers have shifted from equity-heavy funding to bond-heavy funding as cash needs outstripped operating cash flow. The private credit markets are already showing strain: spreads are widening on lower-rated borrowers, and institutional investors are pulling back. If the AI buildout fails to produce returns that can service that debt, lenders take losses. Those lenders include pension funds, insurance companies, and sovereign wealth funds. Even if the debt is not on bank books, the BIS notes, repo markets and prime brokerage lines connect the AI sector to the core liquidity layer of the financial system.
Ed Zitron, a frequent critic of AI hype, described the delayed IPOs of OpenAI and Anthropic as symptoms of a sector that cannot meet its funding needs. The combination of insatiable capital hunger and zero near-term return prospects, he argued, makes a bad ending baked in. The incentives to keep the party going, however, remain massive. Fund managers who sit out the AI trade risk underperforming benchmarks. Companies that stop spending risk ceding the narrative to competitors.
The BIS does not forecast a crash. It warns that the financial system is less resilient to one than markets assume. Margin debt near record highs, circular financing, debt dependence, and household equity exposure create a risk stack that no single shock would need to trigger. A disappointing earnings report from one hyperscaler, a credit downgrade, a rising cost of debt – any of these could start the unwinding.
The BIS’s own historical analysis, published alongside the report, shows that when margin debt reaches current levels, subsequent corrections take longer to recover from. The 1929 crash took 25 years. The dot-com bust took 15 years for the Nasdaq to reclaim its peak. The BIS does not forecast the timeline. It simply notes the numbers.
For context on how central bank warnings have played out historically, read our broader market analysis.
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.