What it is
Circular financing describes a set of deal structures now common among the largest AI infrastructure companies, in which the same dollar functions as both an investment and a sale. A chipmaker or cloud provider invests capital in a customer; that customer then uses some of the capital to buy the investor’s products or services. IDC, the technology research firm, defines it as “investment structures where the same capital flows simultaneously as vendor payment and equity stake” — money that circulates within a small group of firms rather than entering as independent, arm’s-length market demand.
The pattern is not new to corporate finance. Commentators have compared it to “round-tripping,” a practice associated with the dot-com era in which companies traded revenue with each other to inflate reported growth. What distinguishes the current wave is scale and interconnection. A widely cited example: Nvidia commits to invest in OpenAI; OpenAI buys cloud capacity from Oracle; Oracle buys chips from Nvidia to build that capacity; Nvidia also holds an equity stake in CoreWeave, a cloud provider that leases capacity to OpenAI. Money and product move around a closed loop involving a handful of companies, each appearing in more than one role — investor, supplier, customer — in relation to the others.
Analysts disagree about how alarming this is. Defenders argue the arrangements reflect ordinary supply-chain economics under chip scarcity: building AI data centers is capital-intensive, the most advanced chips are supply-constrained, and pairing long-term purchase commitments with financing is a standard way to secure both supply and revenue visibility. Critics counter that circular structures can obscure whether reported revenue reflects genuine end-user demand or self-generated internal transactions, and that they concentrate risk — if demand for AI services falls short of expectations, losses could cascade through several companies at once rather than being contained to one.
Why it matters for AI governance and narratives
Circular financing sits at the intersection of two contests the observatory tracks: the framing battle over whether AI investment reflects durable demand or speculative excess, and the question of who bears the cost if that judgment proves wrong. When a company’s revenue growth is partly a function of investment it received from its own supplier, claims like “record revenue” require unpacking rather than standing as facts. A company’s choice to disclose these structures — rather than let them be inferred from scattered filings — is itself a signal about how confident, or how exposed, it judges itself to be.
The topic also shapes how regulators and financial-stability bodies frame AI risk. Circular-financing arguments have begun appearing in discussions of systemic risk: the concern that a handful of interlocking companies — Nvidia, Microsoft, OpenAI, Oracle, CoreWeave, AMD, among others — have become financially entangled enough that a shock to one could propagate through the rest. That reframes AI infrastructure spending as a matter of financial stability rather than only industrial policy, which changes who has standing to ask questions about it.
Key facts and dates
Several large, publicly reported deals illustrate the pattern. In September 2025, Nvidia announced a letter of intent to invest up to $100 billion in OpenAI, tied to OpenAI deploying at least 10 gigawatts of Nvidia systems; in February 2026, Nvidia committed roughly $30 billion as part of a larger $110 billion OpenAI financing round. Oracle’s cloud agreement with OpenAI, announced in 2025 as part of the “Stargate” project, is reportedly worth around $300 billion over five years and requires Oracle to buy roughly 400,000 Nvidia GB200 chips. OpenAI’s October 2025 partnership with AMD paired a commitment to deploy 6 gigawatts of AMD chips with a warrant for OpenAI to acquire up to roughly 10% of AMD’s shares. OpenAI’s commitments to CoreWeave — a cloud provider in which Nvidia holds a large equity stake — total about $22.4 billion; Nvidia separately pre-purchased $6.3 billion of CoreWeave capacity in September 2025. IDC estimates that Microsoft’s AI-related performance obligations, of which roughly 45% relate to OpenAI, total around $600 billion.
Reporting on the total scale of these arrangements varies by methodology. Some outlets put the value of Nvidia’s circular commitments alone — including a reported $500 billion cooperation agreement with South Korea’s SK Group and up to $250 billion in OpenAI data-center lease guarantees under discussion in mid-2026 — near $750 billion; other analyses of the broader AI supply chain put total circular arrangements above $800 billion. These are order-of-magnitude estimates, not an audited industry total; no single accounting standard currently captures what should count as “circular” financing across the sector.
Where to learn more
- IDC: “Circular Financing Has Muddied the AI Story” — industry-analyst definition and case data
- The Register: “AI’s trillion dollar deal wheel bubbling around Nvidia, OpenAI” — deal-by-deal timeline
- Program on Negotiation, Harvard Law School: “What Are Circular Deals?” — definition and comparison to round-tripping
- Axios: “Nvidia reignites ‘circular’ AI concerns as it weighs OpenAI financing guarantee” — current reporting on Nvidia-OpenAI negotiations