BIS warns debt-fueled AI infrastructure spending threatens global financial stability

The Bank for International Settlements warned on June 28 that debt-fueled AI capital expenditure poses a systemic risk to global financial stability, in its Annual Economic Report 2026. The spending is projected to top $1 trillion across the five largest hyperscalers in 2025-2026.

The BIS, the Basel-based institution often called the central bank for central banks, said the spending is outpacing the hyperscalers’ earnings and free cash flow. That gap is the heart of its concern: when investment runs on borrowed money rather than profits, a downturn hits balance sheets and lenders at once.

The report flagged several risk mechanisms. Circular financing, in which chipmakers and cloud giants take equity in AI labs that then buy their compute, recycles revenue back to the original sellers. Reliance on non-bank financing from hedge funds and private credit draws lighter oversight. Extended depreciation schedules on data-center gear obscure true costs. And surging demand for power and components feeds inflation.

The BIS press release said “optimism surrounding AI may not last, despite its promise of future productivity gains,” warning that “intense competition for market leadership may fuel over-investment, as seen in previous innovation waves.”

Pablo Hernández de Cos, the BIS general manager, said “success depends on sound fiscal and financial foundations.” Frank Smets, acting head of the BIS monetary and economic department, said the new tie between sovereign and financial markets “may mean more frequent and sharper drops in sovereign bond values.”

One widely cited figure — $8 trillion in AI build-out costs over six years — comes from a Columbia University economist estimate quoted in coverage, not from the BIS itself. The report stops short of calling the boom a bubble, framing it instead as overinvestment risk against a historical pattern.