BIS says 55% of AI funding comes from other AI firms

A new Bank for International Settlements bulletin finds that 55.2 percent of AI funding comes from other AI firms, mapping what Crypto Briefing described as a tightly looped financing web in which suppliers often bankroll their own customers¹. MarketScreener summarised the same work as AI being 55 percent financed in a closed loop, and said BIS economists had taken up the question directly³.

The BIS warned of potential macroeconomic risks stemming from circular relationships among AI companies². It also warned that the US pace of investment could end in an “investment bust”⁴.

Why it matters to investors

Circular financing matters because it obscures where the credit risk finally sits. When the firms selling compute are also the firms funding the buyers, a slowdown in end demand does not stay contained to one balance sheet, which is the amplification channel the BIS flags^1,2.

The debt is already dispersed through public markets. The Information reported that AI data centre debt is showing up everywhere, with high-yield bond exchange-traded funds from asset managers including State Street and Charles Schwab counting debt tied to Fluidstack-leased facilities⁵.

Rates compound the exposure. The New York Times reported that one data centre company recently warned that interest payments on its floating-rate debt could rise by about $30 million a quarter for every percentage point of increase⁶.

MarketScreener noted that the question of how AI investment is financed had, until this bulletin, been left largely to commentary rather than to measurement, and that BIS economists have now put a figure on it³. Crypto Briefing’s reading of the bulletin is that the loop runs through suppliers funding customers, which is the specific arrangement that makes a demand slowdown travel back up the chain¹.

What to watch

Europe offers a contrast case. ECB data shows that Europe is not borrowing to fund AI, though Cryptopolitan reported that the lag still worries European economists, who frame the gap as either prudence or handicap⁴.

For investors the near-term signals are the ones that would break the loop from the outside: the behaviour of high-yield funds that hold data centre paper, and the sensitivity of floating-rate borrowers to further increases in long rates^5,6. The BIS bulletin itself is a measurement of the structure, not a prediction of its failure, and the agency’s own framing is of risk rather than of an outcome^1,2.