The Circular Money Machine Behind the AI Boom

A regulatory filing recently revealed that Nvidia holds a roughly $21 billion stake in SpaceX, its second-largest investment in any company. The stake came from the $10 billion Nvidia put into Elon Musk’s xAI before it merged into SpaceX, and Musk has said SpaceX will build its AI only on Nvidia chips. Nvidia, in other words, is a major owner of one of its own biggest customers, and that arrangement is quietly everywhere in the AI boom.

Every quarter, big investors must disclose what they own in a filing called a 13F. Nvidia’s latest one contained a line that stopped analysts cold: a stake in SpaceX worth about $21 billion, the chipmaker’s second-largest holding in any public company. At first glance it reads as a vote of confidence from the smartest chip company on the planet. Look closer and it is something stranger. The position is simply Nvidia’s own $10 billion investment in Musk’s xAI, made in January and marked up after xAI folded into SpaceX. And the company Nvidia now owns a piece of has promised to build its AI using nothing but Nvidia chips.

This is the shape of the entire AI trade in miniature. Nvidia invests in a company. That company uses the money to buy Nvidia chips. The chip sales make Nvidia’s revenue look enormous, which lifts its stock, and the customer’s rising value lifts the stake Nvidia holds. Money makes a full circle and comes back looking like growth. Wall Street has a blunt name for it: circular financing.

How big the circle has grown. This is no one-off. Nvidia has committed close to $50 billion in equity to the AI companies that are also its customers, from a pledge of up to $100 billion to OpenAI last year to stakes in chip and cloud firms across the industry. In a single week last summer, Bloomberg counted $750 billion in new Nvidia deals. Analysts have stopped whispering about it. Bernstein called the OpenAI arrangement a clear source of circular concerns, and Wedbush said Nvidia’s dealmaking fits squarely into the circular investment theme. When a company funds its own customers, the demand for its products can look far larger than the real, independent appetite underneath.

Why it echoes past bubbles. Vendor financing, as economists call it, is legal and sometimes sensible. It also carries an ugly history. In the telecom bust of 2000, equipment makers lent money to the very startups buying their gear, which made demand look bulletproof right up until the startups failed and the loans and the sales vanished together. The worry today rhymes: how much of the AI boom is real end demand from paying customers, and how much is a handful of giants passing the same dollars around a table. Bain, the consulting firm, put a number on the gap. To justify what is being spent on AI infrastructure, the industry needs roughly $2 trillion in annual revenue by 2030. On current trends, Bain sees it arriving about $800 billion short.

Why this reaches your account. You may never buy a share of Nvidia on purpose, and you almost certainly own a great deal of it anyway. Nvidia is among the largest holdings, and often the single biggest, in most S&P 500 index funds, which means it is likely a top position in your 401(k) too. The same handful of AI names that carried the index to record highs this year are the ones tangled in these deals. That concentration has been a gift on the way up, and it becomes a single point of failure if the market ever decides the circle is spinning on its own momentum. That deserves a clear-eyed look at how much of your money now rides on these few names. Cash and short-term Treasuries, meanwhile, still pay more than 4 percent for anyone who wants to trim exposure while the question resolves.