What's the clearest example of a circular AI deal?

Nvidia’s plan to invest up to $100 billion in OpenAI while OpenAI spent much of that money buying Nvidia chips and cloud capacity is the most-cited case; by early 2026 the investment had shrunk to around $30 billion.

Why does the BIS call these deals "poorly disclosed"?

Its June 2026 annual report found that circular financing terms often aren’t made public in detail, and warned of a risk that the same underlying asset — such as a batch of GPUs — could be pledged as collateral more than once across different deals.

Is circular financing illegal?

No — it’s a legal financing structure, similar in spirit to vendor financing used in past infrastructure booms. The concern regulators raise is about transparency and systemic risk, not wrongdoing.

How is this different from an ordinary supplier-customer relationship?

In an ordinary sale, a customer pays with its own independent capital. In a circular deal, the supplier itself provides some of the money or credit its customer uses to buy that supplier’s own products, so the same funds can effectively be counted more than once across different companies’ books.

Could a slowdown in AI spending cause a wider financial crisis?

The BIS says it’s a real risk: because much of this financing runs through non-bank lenders like private credit funds rather than regulated banks, a sharp pullback in AI spending could spread through the financial system faster than a typical banking crisis.

What Is Circular AI Financing — and Why Are Regulators Worried?

Circular AI financing is a deal structure in which a supplier — usually a chipmaker or cloud provider — invests in or lends money to an AI company that then uses some of that money to buy the supplier’s own products or services. The same dollars, and sometimes the same physical hardware, can end up counted as revenue, as an asset, and as an investment gain by more than one company at once, which is why regulators have started calling these arrangements a financial-stability risk rather than just an unusual sales tactic.

How a circular deal works

The clearest example is Nvidia’s relationship with OpenAI. In September 2025, Nvidia said it “intended” to invest up to $100 billion in OpenAI as OpenAI deployed 10 gigawatts of Nvidia computing systems — meaning Nvidia’s own investment was tied directly to OpenAI buying more Nvidia chips. By early 2026 the deal had stalled and shrunk to roughly $30 billion, but the pattern it illustrates is now common across the industry: a chipmaker or hyperscaler takes an equity stake in, or extends credit to, an AI lab or data-center operator, which then commits to a multi-year contract to buy computing capacity or chips from that same investor. Money — and sometimes reported revenue — flows in a loop between a small number of companies rather than coming from a broad base of independent customers.

These arrangements often overlap with the long-term leases covered in our explainer on neoclouds, where a chipmaker or hyperscaler backs a data-center operator’s debt in exchange for that operator hosting the investor’s own AI-lab customers — one more link in the same chain, rather than a separate phenomenon.

Why banks and regulators are worried

The Bank for International Settlements — often described as the central bank for central banks — named AI financing one of the top risks to the global financial system in its June 2026 annual report. It found that “the terms of such deals are typically poorly disclosed, with risks of the same asset being pledged multiple times,” and warned that a growing share of this lending flows through hedge funds and private credit firms rather than regulated banks — intermediaries that operate with far less oversight, which could let a downturn spread faster than a traditional banking crisis.

Critics have compared the pattern to the vendor-financing loops of the dot-com bubble, when telecom equipment makers lent money to internet startups so they could buy that same equipment — inflating both companies’ apparent growth until the startups couldn’t repay their debts. Supporters of today’s AI deals counter that leasing and vendor financing are how every capital-intensive industry, from railroads to telecoms, has funded infrastructure buildouts, and that no single traditional lender could underwrite hundred-billion-dollar chip orders alone.

Why it matters

The stakes are unusually high because of scale: the BIS estimates the five largest hyperscalers will spend more than $1 trillion on AI infrastructure across 2025 and 2026 combined, already outpacing their free cash flow in some cases. If demand for AI products falls short of what these circular contracts assume, losses wouldn’t stay contained to one company — they could ripple through every supplier, lender, and customer tied into the same web of deals, which is exactly the concentration regulators say they can no longer clearly see or measure. It also feeds into how AI labs get valued in the first place: many of the eye-popping figures reported for private AI companies come from secondary-market trades among investors who are themselves betting that these circular commitments will keep paying off.

In the news

Anthropic’s talks to lease up to $10 billion in computing power from Meta show the same underlying pattern spreading to new pairings of companies — read more in our brief on the negotiations.

How a circular deal works

Why banks and regulators are worried

Why it matters

In the news

Frequently asked questions