Circularity

Circularity means that money flows in a circle among a small number of companies. Company A gives Company B money, for example as an investment or a loan. Company B uses exactly this money to buy products from Company A. In the end, Company A has revenue on its books that it effectively paid for itself. The term originates from the world of finance and is usually used as a warning. Since 2024 it has appeared frequently in reports about the major artificial intelligence companies.

Why investors get nervous about round-tripping deals

For investors, revenue is the most important signal that a product is genuinely needed. Strong revenue growth is considered proof of real demand. It is precisely this signal that circular deals make unreliable. That’s because part of the growth then comes not from outside customers, but from the company’s own cash.

The dependency involved is also dangerous. When a small number of companies finance one another, they all hang by the same thread. If one participant runs into trouble, several revenue sources disappear at once. Experts call this a concentration risk: the risk is not spread out, but concentrated within a small circle.

There is a well-known historical precedent for this. Around the year 2000, network equipment makers sold technology to young internet companies and financed the purchases themselves. When the buyers went bankrupt, the equipment makers' revenues collapsed. Anyone talking about circularity today often has this episode in mind.

How the money goes round in a circle

The typical process has three steps. First, a large corporation takes a stake in a smaller company and transfers several billion. Second, this company commits to purchasing services from the investor for a similarly large amount. Third, the corporation books these purchases as ordinary revenue.

In the AI sector, the services purchased are usually computing time or specialized chips. Training large language models requires enormous data centers, and these are rented out by exactly the same corporations that are also investing. Sometimes no cash changes hands at all; instead, chips or server capacity are contributed directly. This is known as an in-kind contribution instead of a payment.

An important distinction: circularity is not automatically fraud. An investor who also sells technology to the company it holds a stake in is acting legally and often sensibly. It becomes problematic when the interconnections are not clearly disclosed in the financial report. Accounting rules require disclosures for so-called related parties, meaning business partners with close ties to one’s own company.

Circularity in AI headlines

The term comes up most often in connection with the chipmaker Nvidia and major AI labs. Nvidia invests billions in companies that subsequently pay for Nvidia chips or data centers built with those chips. Contracts between cloud providers and AI developers are also discussed in this way. Here, cloud simply means computing power that one rents in someone else’s data centers.

As a reader of financial news, you can recognize the pattern from certain phrases. There is talk of a strategic investment paired with a simultaneous supply agreement, or of a customer who is also a shareholder at the same time. Analysts then ask how large this one partner’s share of revenue is. If it is in the double-digit percentage range, that is considered a warning sign.

A common misconception is that circularity makes all figures worthless. That is not true. Part of the demand for computing power comes from completely unrelated customers, such as banks or software companies. The task is to separate the two parts. That is exactly why the term is useful: it helps you ask good questions instead of blindly believing in growth.