The Bank for International Settlements said in its June 2026 Annual Economic Report that the five largest hyperscalers are set to spend more than $1 trillion on AI-related capital expenditure in 2025 and 2026. It said commitments exceed earnings and, for some companies, free cash flow, contributing to higher debt issuance.
The BIS said the spending surge has supported current growth and financial conditions, but warned that if revenue and productivity gains fall short of expectations, financing could be withdrawn quickly. It said the effects could spread through equity prices, credit spreads, investment and household wealth, while valuations of companies central to AI development imply strong long-term earnings growth assumptions and the large share of US stocks in global equity benchmarks increases the scope for international spillovers from a US-led repricing.
The report also highlighted supply constraints in electricity, advanced semiconductors and grid equipment, saying higher computing demand is already putting pressure on power prices and input costs and could spill over into inflation. It added that opacity in financing arrangements adds to vulnerability, citing private agreements among hyperscalers, chip makers and AI labs, including equity stakes tied to multi-year purchase commitments and outsourced data-center structures with long-dated lease obligations and limited disclosure.
In a January 2026 bulletin, the BIS said AI-related investment had become a material driver of US activity, with data centers and IT manufacturing facilities at about 1% of GDP by mid-2025 and broader IT investment at about 5% of GDP, while outstanding private credit loans to AI-related companies had risen to more than $200 billion. A March 2026 BIS review said hyperscaler bond issuance exceeded $100 billion in 2025 and that off-balance-sheet funding vehicles backed by private credit and lease cash flows are increasing links between AI infrastructure spending and non-bank credit markets.