Central bankers grow nervous about AI funding

The Bank for International Settlements published its Annual Economic Report on June 28, warning that the AI boom is becoming a source of financial instability. That alone is not surprising.

What is surprising is the specific mechanism the BIS is worried about: not whether AI pays off, but who will take on the debt if it doesn’t.

The BIS calls itself the central bank for central banks. When it flags a funding structure as a systemic risk, regulators tend to listen.

What is an AI bubble?

An “AI bubble” does not mean AI stops being useful. It means the money funding AI infrastructure grows faster than the cash flow that infrastructure generates, and investors keep paying anyway.

That gap is now measurable. According to a BIS bulletin published this year, AI hyperscalers have ramped up capital spending so sharply that free cash flow has recently lagged capital expenditures in absolute dollar terms.

These firms historically ran with less debt than typical companies, funding growth from their own profits.

That model is breaking. The BIS bulletin states plainly that the sheer scale of AI investment is testing the limits of what cash flow alone can support.

How the financing quietly shifted from cash to debt

When cash flow cannot keep up, companies borrow. Hyperscaler corporate bond issuance topped $100 billion in 2025, according to the BIS Quarterly Review, mostly in long-term debt locking in funding for multi-year data center build-outs.

That number matters because credit default swap spreads on these bonds also rose during the same period, the BIS found, signaling that bond investors are pricing in more uncertainty about whether the projects will pay off.

More Economy:

The bigger shift is what is happening off the balance sheet. Hyperscalers are increasingly using joint ventures and special-purpose vehicles, capitalized by private credit firms, to build data centers without the debt showing up on their own books, per BIS research.

The debt does not disappear just because it is harder to see. Private credit funds originated more than $40 billion in loans to AI-related companies in 2025 alone, according to BIS data, a fivefold jump in a market with far less disclosure than public bond markets.

The BIS warns that AI infrastructure is increasingly financed through private credit and hedge funds, channels with less oversight than banks.

Richard Newstead / Getty Images

The hidden loop of circular financing

A related concern is circular financing, where chip and cloud giants take equity stakes in AI startups that then spend that same capital buying chips or compute from the investor, as Bloomberg has reported.

The risk is that these deals can inflate the appearance of demand. Revenue looks real on paper, but a chunk of it is the investor’s own money moving through one company and back into another.

Why this structure worries central bankers

Banks are regulated, monitored, and required to hold capital against losses. Hedge funds and private credit vehicles are not subject to the same oversight, which is exactly the BIS’s concern.

BIS General Manager Pablo Hernandez de Cos told reporters the message is one of “urgency,” because today’s debt is increasingly financed through what the BIS terms non-bank financial intermediaries.

Frank Smets, the BIS’s acting head of monetary and economic department, separately warned that record sovereign debt combined with leveraged hedge fund activity in bond markets has created what he called a “sovereign-financial stability nexus,” raising the odds of sharp, sudden swings in government bond prices.

Zhang Tao, the BIS’s Asia-Pacific representative, made the comparison explicit in remarks to the South China Morning Post: if AI sentiment turns, the interconnectedness of these non-bank channels could make a correction move faster than the 2008 banking crisis did.

For investors, the takeaway is not that AI demand is fake. It is that the speed of any repricing now depends on plumbing most people never look at.

Here are four numbers worth watching:

Hyperscaler bond issuance crossed $100 billion in 2025, locking in long-term debt against AI infrastructure that has not yet proven its return.
Private credit originations to AI companies exceeded $40 billion in 2025, a market with limited public disclosure compares to bonds.
Credit default swap spreads on hyperscaler debt rose through the review period, a market signal that bond investors are pricing in more risk.
The BIS explicitly urged policymakers to extend oversight beyond traditional banking, a direct nod to private credit’s growing role.

The bigger pattern

Every infrastructure boom eventually outgrows the balance sheets that started it, from railroads to telecom fiber to housing.

What separates this one is the BIS naming the exact channel, non-bank private credit, that previous crises took years to identify after the damage was done.

The central bank of central banks rarely names a specific financing structure as a watch item this early.

The open question is whether regulators move fast enough to add oversight before the debt comes due, or whether they are once again diagnosing a bubble only after the air starts leaking out.