Jim Rickards AI Debt Warning: What's Real

If you Google “AI stocks” right now, you get hype. If you read Jim Rickards’ presentation — and you should, it is free — you get the other side: a 3.5-month campaign arguing that the machinery financing the AI buildout is the most important story the market is not talking about.

This is not a stock pitch. No teasers, no tickers for sale. It is a free presentation from Paradigm Press, and that matters because it changes the incentive. When the product is a subscription, the publisher needs you to buy. When the product is a warning and the presentation is free, the publisher needs you to watch and share. The motivations are different, and as a reader you should weight that.

What this campaign actually is

Since March 18, 2026, Rickards has published at least 13 press releases through GlobeNewswire, each one adding a new layer to the same thesis: the AI infrastructure buildout is being financed through debt structures that most investors do not understand and are not being told about.

The releases read like a serialized investigation. March introduced the idea. April went deeper — off-balance-sheet vehicles, circular financing, insider selling. By June the campaign had a date: July 29, when Meta and other AI-linked companies report earnings.

This is pattern is intentional. Rickards is not selling a one-off prediction. He is building a case, piece by piece, with receipts, and the July 29 earnings cycle is the climax.

The $200 billion number

The headline number comes from the June 22 release: AI companies and projects raised at least $200 billion through debt markets in 2025. That is investment-grade bonds, high-yield debt, and private credit — much of it sitting outside the public markets retail investors track.

Rickards is not making this up. He cites JPMorgan analysts who have noted that bond portfolios increasingly track tech company performance rather than traditional interest-rate dynamics. He cites Oliver Wyman, which has warned that lenders may hold more data-center exposure than their internal models capture.

The argument is simple: this debt exists, it is large, and most of the people who own it through pension funds and retirement accounts do not know they own it.

The Enron and Lehman comparisons

This is where the campaign gets its teeth. Rickards argues that major AI companies are using special purpose vehicles (SPVs) to finance data centers — keeping debt off their primary balance sheets just like Enron and Lehman did before their collapses.

He also points to “circular financing” — companies lending each other money to buy each other’s products, generating reported revenue while the underlying economic exchange creates little real value. He compares this directly to Cisco’s sales practices during the dotcom boom.

The Enron comparison is the strongest claim in the presentation, and it is worth a skeptical read. Rickards states flatly: “These were the same shady accounting practices used by Enron and Lehman Brothers.”

Is that fair? Enron was fraud — active, documented fraud. An SPV used for legitimate financing is not the same thing as an SPV used to hide losses. But Rickards’ point is not that the accounting is identical. His point is that opacity in financing structures creates the conditions for a crisis, whether or not anyone is breaking the law. That is a softer claim and a fairer one.

The insider selling data

The campaign’s most detailed release — April 6 — documents a specific pattern: AI company insiders have liquidated roughly $1 billion in shares, not in a panic but in a methodical, deliberate pattern. Rickards argues this reflects knowledge that has not yet reached public markets.

This is a concrete, verifiable claim. Insider selling data is public. It can be checked. If true, it is the strongest signal in the entire presentation — because insider behavior, across history, has been one of the most reliable indicators of what happens next.

Why July 29 matters

Meta is expected to report earnings around July 29. Rickards uses this date as the campaign’s centerpiece: if Meta (or another major AI-linked company) delivers growth forecasts or spending projections below expectations, he argues it could function as a “dotcom moment” — a single earnings miss that confirms the thesis and triggers a broader repricing.

The dotcom comparison is useful. In early 2000, a handful of earnings misses from internet companies confirmed what skeptics had been saying: the cash was running out. The Nasdaq eventually fell nearly 80%. Rickards believes a similar dynamic could play out if AI companies fail to deliver on the spending and growth assumptions baked into current valuations.

What to make of it

Three observations from someone who reads these things for a living.

First, this campaign is unusually thorough for a free presentation. The press release cadence — 13 releases over 105 days — suggests a deliberate, well-resourced production. Paradigm Press is treating this as a flagship effort.

Second, Rickards’ track record deserves honest weight. He warned about 2008 two years in advance. He called the COVID crash three weeks early. He was early on Trump and Brexit when the consensus was wrong. That does not make him infallible — he has made bearish calls that did not materialize on his timeline — but it makes him worth reading, especially when he is giving the analysis away.

Third, the free presentation format changes the incentives. There is no subscription to upsell here. The call to action is to watch a video. That means the conflict-of-interest concern that follows every paid promo — “is this true or are you just selling me a newsletter?” — is significantly reduced. Not eliminated. Reduced.

Your move

The Jim Rickards AI debt warning campaign is the most sustained, well-sourced free analysis this site has seen in its opening month. The July 29 earnings cycle is real. The $200 billion debt figure is real. The insider selling data is public and checkable. The structural comparisons to 2008 and Enron are aggressive — but the underlying concern about opacity in AI financing is legitimate.

You can watch the free presentation yourself. Then, on July 29, watch Meta’s earnings the way you would watch a weather report when a storm warning is posted. You do not have to act. You just have to pay attention.

If Rickards is wrong, you have lost nothing. If he is right, you had the date in your calendar.