Standing on One Leg: The IMF Puts AI’s Circular Financing Above $1 Trillion

Kristalina Georgieva put a number on it. Speaking on Nov 16 at an event tied to the Association of Southeast Asian Nations, the managing director of the International Monetary Fund said circular financing among artificial-intelligence companies has grown to slightly more than $1 trillion, and that about two-thirds of it has no matching external revenue behind it.

She described the structure as standing on one leg. The image is exact. A company financed by its own customers, whose customers are financed in turn by the company, has one source of support where an ordinary business has two: money from outside the circle.

Investors have watched these arrangements for a year without being able to size them. The deals are private, the disclosures are thin, and each one is defensible when examined alone. The fund’s contribution is aggregation. One trillion dollars with two-thirds unsupported is a different fact from a sequence of press releases.

The mechanism is old: suppliers have financed buyers in capital-intensive industries for as long as there have been suppliers. What is new is the concentration. A handful of firms sell the chips, rent the computing, invest in the buyers and purchase services from them. A dollar that moves inside that group can appear as revenue on one set of books and as an investment on another.

Georgieva’s second figure reaches beyond the technology industry. Hyperscale cloud providers now account for roughly one-tenth of all investment-grade bond issuance worldwide, up from less than 1% in 2024. In two years a group of technology companies has become one of the largest borrowers in the global credit market.

That share matters because of how bond indices work. Once an issuer is large enough it enters the benchmarks that funds track, and funds buy it because it is in the index rather than because they have formed a view on it. A pension fund in Asia or an insurer in Europe ends up holding data-centre credit without anyone having decided to buy it.

The venue of the warning is worth noting. Georgieva spoke in a region that has spent two years courting data-centre investment, and whose governments have been offered computing capacity, training and capital in exchange for land, power and fast permits. Her audience was not Silicon Valley. It was the finance ministries being asked to underwrite the buildout.

Emerging markets carry a specific exposure. Borrowing in dollars to finance capacity whose revenue arrives in local currency creates a mismatch that has nothing to do with AI and everything to do with how the past three decades of currency crises began. Several governments have guaranteed power contracts and granted tax abatements without booking the contingent liability.

The fund published no estimate of losses, and it did not name the companies it was describing. The restraint is deliberate. An institution that names firms in a stability warning moves markets and invites litigation. An institution that describes a structure can be argued with.

A fair defence of the structure exists and deserves stating. Circular deals let a company with a good idea and no balance sheet obtain computing it could not otherwise afford. Vendor financing built the telecommunications industry, the aircraft industry and much of the semiconductor industry. Most of the arrangements now being scrutinized will be repaid in full.

The problem is the tail. Two-thirds of $1 trillion has no external revenue behind it, which means repayment depends on the AI market continuing to grow at the rate it has grown. One leg holds the weight perfectly well until the weight shifts. Then the question is not whether the structure was reasonable but who is standing behind it.

That question has an answer in every financial crisis, and the answer is rarely the party that signed the contract. Depositors, policyholders and taxpayers absorb the loss because the entity that took the risk was too interconnected to fail and too large to ignore. The fund’s role is to say this before it has to lend. Georgieva’s remarks are the earliest form of that warning: a figure, an image and a venue chosen to reach the people who would have to pay.

What follows is measurable. If the fund repeats the number at a larger size in its next stability report, the structure grew. If it publishes a distribution of maturities and counterparties, the supervisors found a way to look inside. Until then the useful question for any investor is where an AI revenue line came from, and whether the money started inside the same group.

The distinction the fund is drawing is between revenue earned and revenue recycled. Both appear on an income statement, and only one of them can be spent outside the family. A dollar that goes around the circle twice is counted twice and supports nothing.

There is a political dimension that will decide how quickly any of this is addressed. AI investment is currently the main source of good news in several large economies, and the officials being asked to slow it down are the same officials who advertise it. The fund can describe the risk with precision. It cannot make a government act on the description while the buildout is still creating jobs in its constituency.

Georgieva chose her language carefully for that reason. A structure standing on one leg is not a prediction that it will fall. It is a statement that the number of ways it can fall has been reduced to one, and that nobody has yet said what happens on the day it does.

Related Posts

  • November 20, 2026
  • 13 views
Sandisk Completes Eight-for-One Split as NAND Outlook Darkens

Eight for one. Sandisk’s stock split took effect on Nov. 19, and the shares began trading on the adjusted basis on Nov. 20, a change that divides the price by…

  • November 20, 2026
  • 8 views
Qualcomm Cuts About 500 Jobs in California While It Pushes Into AI

The confirmation came on Nov 20, in the flat language companies use for this kind of news: about 500 positions eliminated in California, spread across departments in San Diego and…