AI debt boom raises financial stability concerns, BIS says

What happened

The Bank for International Settlements said AI infrastructure is increasingly funded by long-term debt and complex off-balance-sheet arrangements, which means financing kept off the main balance sheet. Capital spending by the five largest U.S. hyperscalers is expected to exceed $800B by the end of 2026, with gross bond issuance reaching $222B by September.

Confirmed

Global impact / market context

Heavy borrowing spread across only a few large companies ties their spending plans to bond investors. If AI cash flows disappoint, those companies may need to slow data center building, which would hit equipment suppliers and software vendors that depend on the same spending.

Analyst inference

The BIS said spillovers into broader bond markets have been limited so far. Still, continued borrowing at the current pace could pressure long-term yields, which are the interest rates on long-dated government and corporate debt, raising borrowing costs for companies outside the AI sector too.

Analyst inference

What to watch

1. Watch whether the five largest U.S. hyperscalers keep capital spending on track to exceed $800B by the end of 2026, and whether bond issuance continues at the $222B pace reached by September. Confirmed
2. Track how much borrowing moves through off-balance-sheet vehicles, which the BIS puts at $75B to $100B in linked bonds and asset-backed securities this year, since these debts can stay less visible to investors. Proposed
3. Monitor long-term Treasury yields and data center lease commitments, which Moody's figures put at $1.2T by mid-2026 including $820B not yet started, because higher yields could lift financing costs across the economy. Analyst inference

Evidence

• seekingalpha.com