The AI Bubble Debate: Who's Warning and Why

Jim Rickards isn’t alone in warning about an AI bubble. He’s one voice in a growing chorus.

That’s the story here. Not who’s right. Not who’s wrong. But who’s saying what, how they overlap, and where they diverge. Because when smart people approaching the same question from different angles all arrive at similar conclusions, you should pay attention — even if they disagree on the mechanism.

Here’s what I’ve found mapping the conversation.

The Financial Architecture Camp

This is the camp Rickards leads. The argument: AI companies are financing the biggest capital buildout in history with borrowed money, and the debt structures supporting it look fragile.

Rickards dropped his first presentation in April 2026, laying out how off-balance-sheet SPVs and circular financing schemes in the AI sector mirror the pre-2008 playbook at Enron and Lehman. His second presentation in June put a number on it: $200 billion in AI-related debt raised in 2025 alone. He’s watching July 29 — the next big earnings cycle for AI-linked companies — as a potential stress test for the whole financing structure.

Then there’s the evidence from Wall Street itself.

JPMorgan launched a credit default swap basket in March 2026 covering Alphabet, Amazon, Meta, Microsoft, and Oracle. Let that sink in: the biggest bank in America created a product specifically to let investors hedge against AI companies defaulting on their debt. JPMorgan’s own analysts estimate $40 billion to $150 billion in leveraged loans inside CLOs face disruption from AI-driven shifts. The five hyperscalers the bank covered issued $121 billion in bonds in 2025 — up from a $28 billion annual average between 2020 and 2024.

Morgan Stanley’s research tells the same story from a different angle. AI-related global debt issuance is on pace to exceed $570 billion in 2026 — more than double the prior year. Their analysts note that hyperscaler capex is approaching levels that consume nearly 100% of operating cash flow, compared to a historical average of roughly 40%. When the credit markets are the difference between keeping the lights on and collapse, the question isn’t whether you believe in AI — it’s whether the banks still believe in you.

Garrett Baldwin over at the Money Printer called it first: “The Big Short 2.0.” JPMorgan packaging CDS on AI debt is a signal. Banks don’t build hedging products for risks they think are imaginary.

The financial architecture camp says: the debt breaks first.

The Labor Displacement Camp

Michael Burry isn’t making a debt argument. He’s making a people argument.

In February 2026, Burry amplified a piece from Citrini Research titled “The 2028 Global Intelligence Crisis” with a simple line: “And you think I’m bearish.” The Citrini report — a thought experiment set in June 2028 — imagined a world where AI-driven white-collar unemployment hits 10.2%, the S&P 500 falls 38% from its highs, and an economic feedback loop with “no natural brake” sets in.

The piece went viral. And I mean went viral — it triggered an actual sell-off. IBM dropped 13% in a single day, its worst since 2000. DoorDash, AmEx, KKR, Blackstone all fell 6% or more. The software sector lost nearly $2 trillion in market cap in the weeks that followed. Wall Street dubbed it the “SaaSpocalypse.”

Citrini’s argument is specific: initial AI adoption boosts margins and earnings as companies replace workers with machines. Headline numbers look great. But beneath the surface, real wage growth collapses, white-collar workers lose their jobs, consumer spending contracts, and eventually the whole economy catches a cold that turns into pneumonia. The AI economy keeps improving. The broader economy deteriorates. Those two curves eventually intersect, and it’s not pretty.

Burry has been warning about this for a while. He’s flagged parallels to the 2000 dot-com buildout, slowing cloud growth, widening capex gaps, and what happens when massive data-center spending drains cash flow and forces Big Tech to rely on debt or creative accounting.

The labor displacement camp says: employment breaks first.

The Liquidity Camp

Then there’s the argument that it’s not about debt or jobs at all — it’s about where all the money is going.

HTX Research published a report in mid-2026 titled “AI Bubble Liquidity Black Hole” that lays this out clearly. Their thesis: from late 2024 through mid-2026, a massive share of newly created dollar liquidity was absorbed by the AI value chain. Equity investors bought AI stocks. Bond investors bought AI credit. Private funds financed data centers. Banks lent to tech companies. And everything else — crypto, small caps, emerging markets — got squeezed.

They call AI a “liquidity black hole” for global risk capital. The marginal dollar doesn’t flow to Bitcoin or commodities or real estate. It flows to Nvidia, OpenAI, data-center debt. And if that flow reverses? Everything that was starved of capital by AI’s gravity well doesn’t automatically rebound — it first goes through a painful repricing.

ZeroHedge and the Mises Institute made a related argument under the banner “The Real Threat Is Artificial Credit, Not Artificial Intelligence.” From an Austrian economics perspective, the AI boom is a textbook case of malinvestment driven by distorted interest rates. The Fed’s prolonged monetary expansion created cheap credit, which directed capital into long-horizon, capital-intensive AI projects that can only survive as long as the cheap money keeps flowing. When it stops, the malinvestment gets exposed.

The liquidity camp says: capital flows break first.

The Bankruptcy Camp

Nassim Taleb’s warning is the simplest and maybe the scariest.

He’s not making a macro argument about debt structures or employment or liquidity. He’s saying something more direct: specific software companies will go bankrupt because AI makes their products obsolete.

Taleb went on Bloomberg in February 2026 — right as the Citrini panic was unfolding — and made the point bluntly. When asked if he expects bankruptcies in the software space, his answer was one word: “Definitely.”

His logic is straightforward. The recent rally was driven by a small number of names. AI leaders will not necessarily be the ones that profit from AI — history shows pioneers of technological revolutions almost always get replaced. Tail risk across the sector is structurally underpriced. The danger isn’t a small correction. It’s a large drawdown.

Taleb isn’t predicting a systemic collapse. He’s predicting company-specific failures. Palantir down 22% year-to-date. ServiceNow down 31%. IBM’s worst day in 25 years. The software names that benefited from the old world get killed by the new one before the new winners emerge. That’s not a macro crisis — it’s a business-model singularity. But the result for investors holding those stocks is the same.

The bankruptcy camp says: business models break first.

Where They Overlap

All four camps agree on the same three facts:

1. AI spending is enormous. No one disputes this. $200 billion in AI debt in 2025. $570 billion projected in 2026. Hyperscaler capex consuming nearly all operating cash flow. The numbers are not in dispute.

2. Returns are unproven. The revenue coming in doesn’t justify the spending going out. Every camp acknowledges this gap. It’s the prerequisite for every argument in this conversation.

3. The financial structure is fragile. Whether you’re worried about debt defaults, employment spirals, liquidity vacuums, or business-model obsolescence — everyone agrees the current setup is vulnerable to a trigger.

Where They Diverge

The disagreement is about mechanism and timing.

Rickards says debt structures break first — the off-balance-sheet SPVs, the circular financing, the credit market repricing. Watch July 29.

Burry and Citrini say employment breaks first — white-collar displacement collapses consumption, which collapses the economy, which collapses markets. Watch the labor data.

HTX and the Austrians say liquidity breaks first — AI absorbs so much capital that the rest of the economy starves, and when the flow slows, everything reprices at once. Watch the credit spreads.

Taleb says business models break first — the software companies of today get killed by the AI of tomorrow, one by one, before anyone notices the macro picture. Watch the bankruptcies.

Same destination. Different roads.

The Honest Take

When Jim Rickards, Michael Burry, Nassim Taleb, JPMorgan, Morgan Stanley, HTX Research, ZeroHedge, and Garrett Baldwin all look at the same thing and come away worried — even if they disagree on the specific reason — the thing they’re looking at is probably real.

You can argue about timing. You can argue about magnitude. But the AI bubble debate isn’t about whether risk exists. It’s about which fuse lights first, and how fast it burns.

That’s the honest picture of this conversation. Read the details. Watch the evidence. Decide for yourself.

For more: Check out the steel man of the AI debt thesis here, the full breakdown of Burry’s warning here, and Rickards’ AI black paper promo coverage here.

— Sarge