Vendor Financing Loops
The circular financing structure that emerged by late 2025 in the AI industry would have particularly interested Kindleberger, who spent decades documenting the self-referential credit mechanisms that characterize the mature phase of manias. The structure was described in financial commentary with unusual bluntness: Nvidia finances OpenAI, which purchases computing power from Oracle, which orders chips from Nvidia — a loop in which each participant's revenue is, in significant part, another participant's expenditure, funded by capital raised on revenue projections that depend on the continuation of the loop. Commentators described it as 'far beyond a pyramid scheme; a perfect financing circle, all based on a sea of debt.'
The Infrastructure Imperative — Contrarian ^ Opus
There is a parallel reading that begins from the material requirements of artificial intelligence rather than its financial structures. The vendor financing loops that Kindleberger's framework highlights as pathological may actually represent the rational response to an unprecedented infrastructural challenge. Training and operating frontier AI models requires compute clusters that cost billions to construct and hundreds of millions annually to operate. The energy requirements alone — approaching the consumption of small nations — necessitate dedicated power plants and transmission infrastructure. This is not speculative excess but physical necessity.
The circular financing between Nvidia, OpenAI, and Oracle reflects the fundamental interdependence of hardware manufacturers, model developers, and cloud providers in creating an entirely new computational substrate for intelligence. Historical analogies to the South Sea Bubble miss the crucial difference: those ventures produced no lasting infrastructure, while these loops are constructing the physical and software foundations for artificial general intelligence. The railroad manias of the 19th century offer a better parallel — they too featured circular financing, spectacular crashes, and accusations of pyramid schemes. Yet they left behind the rail networks that enabled industrial civilization. The question is not whether these loops represent sustainable business models in conventional terms, but whether they can survive long enough to complete the infrastructure that renders them obsolete. From this vantage point, the external capital flowing into these loops is not being deceived by circular revenues but is making a calculated bet that the infrastructure being built will generate returns that dwarf any interim losses from financial engineering.
In the AI Story
Kindleberger documented structurally identical mechanisms in the South Sea Company's share-buy-back schemes of the 1720s, in the cross-holdings of Japanese keiretsu before the 1990 crash, and in the SPV structures of the pre-2008 credit bubble. The sophistication of the mechanism bears no relationship to its sustainability. Each instance shared the feature that Kindleberger identified as characteristic: the revenue generated by the loop is real in an accounting sense but circular in an economic sense. Money flows between participants, generating reported revenue at each step, but the ultimate source of that money is external capital raised on the assumption that the reported revenue indicates genuine economic activity.
The structure is not, in itself, fraudulent — it is a standard feature of immature technology markets where participants are simultaneously customers and suppliers. The distinction between healthy vendor relationships and pathological financing loops is a matter of degree, not kind. At some volume, the circular component overwhelms the non-circular component, and the external capital required to sustain the loop exceeds the external revenue that eventually validates it. The critical question is whether the loop can transition to self-sustaining external revenue before the external capital that sustains it becomes unavailable.
The regulatory response that Kindleberger's framework implies is transparency. When the revenue supporting a company's valuation is substantially composed of purchases by other companies within the same financing circle, this fact should be disclosed with the same rigor that financial institutions disclose counterparty exposures. The circular financing of the AI industry represents a systemic risk invisible to outsiders unless disclosure requirements make it visible — and the failure to disclose it perpetuates the information asymmetry between insiders and outsiders that Kindleberger identified as the primary mechanism through which outsiders are disadvantaged.
Origin
The structure was documented in financial commentary throughout late 2025, with specific reference to the Nvidia-OpenAI-Oracle loop and related arrangements involving Microsoft, Amazon, and the major AI platform companies.
Key Ideas
Historical precedent. South Sea Company, Japanese keiretsu, pre-2008 SPVs — the structure recurs across three centuries.
Not fraud but scale. Healthy vendor relationships become pathological when circular volume dominates.
External capital dependency. The loop requires ongoing external capital raised on projections the loop itself generates.
Transparency as remedy. Disclosure requirements would reduce the information asymmetry.
Appears in the Orange Pill Cycle
Infrastructure Manias and Necessary Bubbles — Arbitrator ^ Opus
The tension between Kindleberger's framework and the infrastructure imperative reveals different temporal horizons for evaluation. If we ask "will these financing structures survive the next 24 months?" Kindleberger's analysis dominates (80%) — the circular dependencies create genuine systemic risk that becomes acute when external capital tightens. But if we ask "what physical and computational infrastructure will exist in 2035?" the contrarian view gains weight (70%) — the comparison to railroad manias suggests these loops, however unsustainable, may be the mechanism through which transformative infrastructure gets built.
The critical synthesis emerges when we examine the relationship between financial structure and technological progress. Kindleberger is entirely correct (100%) that the opacity of these arrangements creates information asymmetries that disadvantage outsiders — the call for transparency is unassailable. Yet the contrarian insight about infrastructure requirements is equally valid (100%) — frontier AI genuinely requires capital expenditures that dwarf any previous information technology. The vendor financing loops are simultaneously a warning sign of speculative excess and the only available mechanism for coordinating investments at the required scale.
Perhaps the most productive frame acknowledges both the pathological and generative aspects of these structures. Carlota Perez's work on technological revolutions suggests that financial bubbles and transformative infrastructure are not merely correlated but causally linked — the bubble provides the irrational exuberance necessary to fund rational infrastructure. From this perspective, the vendor financing loops represent what we might call "necessary manias" — financial structures that are unsustainable by design but essential for their side effects. The regulatory challenge is not preventing these loops but ensuring their collapse doesn't destroy the infrastructure they're creating.
Further reading
- Charles P. Kindleberger, Manias, Panics, and Crashes
- John Carswell, The South Sea Bubble
- Hyman Minsky, Stabilizing an Unstable Economy
- Simon Johnson, analyses of AI capital structures