The AI Money Loop That Looks Exactly Like 1999

You probably saw the headline and read it as strength. Nvidia has now put something like $100 billion into the very companies that turn around and spend it on Nvidia chips. Most people in finance called that a vote of confidence. I read it differently — it is one of the oldest patterns near the top of a boom, and we have watched it play out before.

Nvidia’s own equity bets in the customers that buy its chips — CNBC, 2026.

Inside this issue: how the loop actually works, the last time it inflated a market this size, and the one thing I would want to see before I believed the demand was real.

My read has not changed. We are late in this cycle — year 14 of the 18.6-year clock, asset prices near record highs, credit spreads about as calm as they get. This is the stage where money tends to start moving in circles to keep the story going.

Here it is in three moves. One: a chip maker sells its gear to an AI company. Two: instead of waiting for real, paying customers, the chip maker hands that same AI company the money — an investment, a loan, a compute deal — to go buy more gear. Three: that spending lands as the chip maker’s revenue, the stock climbs, and everyone points at the revenue as proof the demand is booming.

Picture a bar owner who lends you the cash to buy his drinks. His sales look incredible — right up until the night you cannot cover the tab. The demand was never really yours. It was his own money, doing a lap.

The AI money loop, mapped against the 1999 telecom bust. What to notice: much of the “demand” is the sellers’ own money.

This is not new. In 1999 and 2000, Lucent and Nortel were the Nvidias of their boom. They did not wait for the telecom start-ups to find customers — they financed them. Lucent carried about $8.1 billion in vendor-financing commitments, roughly a quarter of its revenue. By early 2001, Nortel was financing as much as 130% of an equipment order — more than the gear even cost — often with nothing securing the loan.

Then the loop broke. Look at the bad loans on their books over a single year, end of 2000 to end of 2001.

What to notice: when the loop broke, the losses did not trickle — they jumped. Source: Newsweek; FCC testimony to the U.S. Senate, Jul 2002 (bad loans as % of total).

The FCC chairman later told the Senate the industry owed around a trillion dollars “much of which will never be repaid.” The Nasdaq fell about 78% from its peak.

Now put today beside it. More than $800 billion of these interlinked AI deals are stacked across the supply chain. OpenAI — the demand sitting at the center of the whole story — is on track to lose around $14 billion this year while promising $100 billion of revenue years from now. The money is moving in the same circle. What has changed is the logos and the size.

What to notice: the story’s core customer loses money today and books the payoff years out. Source: public reporting, 2026.

I will be straight with you, because it matters here. Everyone looks like a genius in a bull market — including me. I started investing at 17, I have lived through one real crash and one smaller one, and the main thing that taught me is to distrust how smart everyone feels right at the top. Circular financing is the market feeling smart. That is exactly when I get more careful, not less.

So every Tuesday morning, before I write to you, I pull the credit and liquidity readings in Smart X Terminal and ask one question: is this boom being funded by customers, or by the sellers themselves? Right now, more of it than I would like is the sellers.

What I cannot tell you is the week it snaps. A loop like this can run longer than it should — cheap money and a good story can keep it spinning another year, maybe two. The direction I am confident about. The timing, nobody owns — me included.

The readings here are as of 26 July 2026, when I wrote this — markets move, so for the live picture I keep Smart X Terminal updated in real time.

Below, for paid readers: the three Smart X Terminal readings I am watching for the moment this loop starts to reverse — the first crack, not the crash — and exactly how I have my own money positioned around an AI top I think is closer than the crowd does.