Find this research useful? Add aibubblequestion.com to your Google Preferred Sources, one click, affects only your own results.

The AI Bubble Question
All pages

The Mechanism

Circular Financing, Explained

Imagine a shop that lends its customers money to buy its own products, then reports booming sales. That, in different costumes, is the mechanism that turned several ordinary booms into historic busts.

Updated

Circular financingWhen companies in the same boom fund each other's purchases, so reported demand is partly recycled money. Like a shop lending customers money to buy its own products.Full definition in the glossary is when participants in a boom fund each other's demand. The sales are real transactions and the revenue is legally booked, but part of the demand originates inside the loop, from the sellers' own capital, rather than from independent customers. Across two centuries the research corpus finds the same two effects every time: the loop extends the boom roughly 2 to 4 years beyond what organic demand would support, and then makes the collapse faster and deeper, because the same linkages that recycled money on the way up transmit losses on the way down. Here are the five documented versions, oldest costume first.

1. Railway share calls (1840s)

In the Railway Mania, investors bought partly-paid shares: a small deposit now, the balance promised for later. Many funded new subscriptions by borrowing against, or selling, other partly-paid shares, stacking promises on promises. The structure let the mania run about two extra years; when companies called in the promised capital during the 1847 credit crunch, defaults revealed that much of it had never existed, and forced selling turned a correction into a rout with shares down 50-70% (Odlyzko, MPRA). The exposure event was not an accounting scandal but a liquidity test: a demand for cash that the paper structure could not meet.

2. Trusts holding trusts (1920s)

The investment trusts of the late 1920s bought stocks with borrowed money, and increasingly bought each other: Trust A owned Trust B owned Trust C, with sponsoring banks underwriting all three. Over $1 billion in new trust securities were issued in the first seven months of 1929 alone (New York Times, 1929). The pyramid supported prices an estimated 2-3 years past organic demand; when the market broke, the layered leverage forced liquidation into the fall, and levered trusts commonly lost more than 90%, worse than the Dow's already catastrophic 89% (Novel Investor).

3. Keiretsu cross-shareholding (1980s Japan)

In bubble-era Japan, banks and companies held more than half of core firms' shares in permanent friendly hands, and banks lent against those inflating shares as collateralAn asset pledged to a lender so that if the borrower cannot pay, the lender can seize and sell it. Loans against fast-depreciating collateral are riskier.Full definition in the glossary. Mutual ownership muted selling pressure and disguised risk for 3-5 years; then, after 1989, it transmitted losses through every balance sheet simultaneously and slowed their recognition for a decade (IDE-JETRO). Japan's drawdown was comparable to other great crashes; its 34-year recovery was not. The circularity is why.

4. Telecom vendor financing (1998-2001)

The canonical case, covered fully on the Telecom & Fiber page. Equipment makers lent customers the purchase price of their own products: $25.6 billion outstanding across nine vendors by end-1999, 30-40% of it judged "at risk," and equal to 123% of the North American vendors' pretax earnings (CNET/McKinsey). Vendor financingA supplier lending customers the money to buy its own products. Inflates the supplier's sales until the customers can't pay.Full definition in the glossary bought the boom an extra 1-2 years. The exposure event was disclosure: Lucent's 2000 write-downs (a $501 million charge, about 41% of that year's earnings) told the market that reported growth had been partly self-funded. The heaviest users of the mechanism fell 90-99%, versus 78% for the Nasdaq overall.

5. Enron's round-trips (1997-2001)

Enron ran the mechanism as outright cosmetics: "round-trip" trades that bought and sold the same asset at matched prices so both sides could book revenue, and "dark fiber swaps" that exchanged unused capacity purely to manufacture sales, run through special purpose vehiclesA separate legal company created to hold one project and its debt, keeping both off the parent company's books.Full definition in the glossary that kept the debt off the books (ZDNet). When consolidation rules forced the SPVs into the open in late 2001, credibility broke all at once: peak above $90 to zero in about a year, faster and more total than any honest company in the same bear market. The lesson: when revenue is circular, there is no floor once the story breaks.

What the five cases have in common

  • The loop extends the boom 2-4 years. Rail calls bought ~2 years; trusts 2-3; keiretsu 3-5; vendor finance 1-2; Enron's trades 3-4. Booms with circular structures outlive their organic fuel.
  • The end arrives as a disclosure or a liquidity event. A write-down (Lucent), a cash call (railways), a forced consolidation (Enron), a margin spiral (trusts), or slow-motion loss recognition (Japan). Something forces the loop's internal money to be counted honestly.
  • The bust is faster and deeper than non-circular busts. Roughly 20-30 percentage points more drawdown in the telecom names versus broader tech; total wipeouts in the levered trusts; three decades of stagnation in Japan.

The pattern is also visible while it is still forming, not only after the fact. On 7 August 2026, CoreWeave closed a $2.6 billion delayed-draw term loan (DDTL 5.5), and the borrower is not CoreWeave, Inc.: it is CoreWeave Financing DDTL V-V, LLC, a Delaware special-purpose subsidiary, secured by that subsidiary's own assets and a pledge of 100% of its equity, maturing 1 September 2031 (CoreWeave, Form 8-K, filed 10 Aug 2026). That is the same structure as the five cases above: a named legal entity, ring-fenced from its parent, borrowing against future contracted revenue. It says nothing on its own about whether this loop will run the course of Enron's or of Meta's audited Hyperion venture, covered on Shadow Debt; it says only that the mechanism above is the mechanism being used now, and that it can be watched rather than reconstructed after the fact.

The reason is mechanical, not moral. Circular structures concentrate fragility at specific chokepoints: the moment a receivable must be written down, a call must be paid, or a cross-held loss must be recognized. Until that moment, reported demand looks strong because the loop is working. The modern question, whether the AI ecosystem's equity stakes, compute commitments, and GPU-backed debt form the same shape, is taken up next on The AI Money Loop, and the defenders' counterarguments get their own page.