The Steel Man: Strongest Case for AI Debt Risk
Five independent signals that make Rickards' AI debt warning hard to ignore
- Chapter
- Promo Literacy
You know what a straw man is. You build a weak version of someone’s argument, knock it down, and declare victory.
A steel man is the opposite. You build the strongest version — the version that survives cross-examination. You find the independent evidence that supports it. You let the argument stand on its firmest ground before you decide whether it holds.
This is the steel man of Jim Rickards’ AI debt thesis.
The core claim: AI companies are financing their infrastructure buildout with staggering amounts of debt. Much of it is off-balance-sheet. Much of it is held by investors who don’t know they’re exposed. When the credit cycle turns — and it always turns — the whole structure could collapse.
Rickards says it at 10,000 feet. Here’s what the data says at ground level.
Independent Evidence #1: JPMorgan Created a CDS Basket on AI Company Debt
The largest bank in America didn’t write a warning. They wrote a trade ticket.
In February 2026, JPMorgan launched a credit default swap basket covering five hyperscalers — Alphabet, Amazon, Meta, Microsoft, and Oracle. Trades in $25 million increments. Five million per name. One-click shorting of the AI debt trade.
This is the same instrument Michael Burry used in 2005 to bet against subprime mortgages. The same structure. The same mechanics. Different decade.
JPMorgan’s own research estimated $40 billion to $150 billion of leveraged loans in CLOs face disruption from AI-driven shifts in the software sector. The bank is simultaneously forecasting that AI will diffuse gradually into the economy AND selling hedging products against an abrupt reset of expectations.
They’re not warning. They’re pricing. That’s what banks do when they see structural risk.
The DTCC confirmed single-name CDS on major US tech firms surged 90% between September and December 2025. A year earlier, CDS on Meta and Alphabet barely traded at all.
Independent Evidence #2: Morgan Stanley Said the AI Supercycle Depends on Credit Markets
The bank that underwrites tech IPOs is saying the quiet part out loud.
Morgan Stanley’s research explicitly frames the AI buildout as a credit story. They estimate $2.9 trillion in AI infrastructure capex through 2028, with a $1.5 trillion financing gap that has to be filled by external capital — corporate bonds, private credit, asset-backed securities, and structured products.
Here’s the part that keeps credit analysts up at night: Morgan Stanley said if credit markets lock these companies out, the AI supercycle ends.
Not slows. Ends.
The bank also warned that software accounts for 16% of the $1.5 trillion US loan market, with 50% of that debt rated B- or lower. More than 80% of software loans are issued by private companies. Nearly 78% are sponsor-backed, meaning their financials are opaque. And 30% of software loans come due by 2028 — versus 22% for the broader market.
That’s a maturity wall. It’s also a refinancing risk. If credit markets get choosy when $235 billion in software loans need to roll over, the companies building AI infrastructure can’t fund themselves.
The whole supercycle sits on a credit-market assumption.
Independent Evidence #3: Michael Burry Shared the Citrini Scenario
February 2026. The guy who called 2008 — who built the trade that became The Big Short — posted a report from Citrini Research on X.
The report laid out a “2028 Global Intelligence Crisis” scenario: rapid AI adoption triggering mass white-collar unemployment, a consumption collapse, and systemic financial distress. It was framed as a thought experiment, not a prediction.
Burry’s caption: “And you think I’m bearish.”
The market reaction was not subtle. The report went viral — 16 million views on X. Software stocks sold off. The Wall Street Journal cited it as a key accelerant of investor anxiety. IBM had its worst day since 2000.
Burry didn’t write the report. He didn’t need to. He just pointed at it and raised an eyebrow. When the most famous contrarian investor in history signals that someone else’s doomsday scenario isn’t crazy enough for him, you pay attention.
Independent Evidence #4: Tech Debt Hit 18% of All Investment-Grade Issuance
This is the structural fact that makes everything else possible.
In 2025, the five largest hyperscalers issued $121 billion in US corporate bonds. Their annual average from 2020 to 2024 was $28 billion. That’s a 4x increase. Wall Street estimates another $100 billion to $300 billion in AI-related bond supply for 2026 alone.
Technology now represents 18% of all USD investment-grade issuance — double the pace from last year. By October 2025, the amount of debt tied to AI had ballooned to $1.2 trillion, making it the largest segment of the investment-grade market. AI debt surpassed US banks as the largest sector in the JPMorgan US Liquid index.
What does that mean? Bond portfolios increasingly track tech performance, not interest rates. When you buy a bond fund, you’re making a bet on AI monetization whether you know it or not.
Oracle alone has $124 billion in borrowings with a Baa2 rating — just two rungs above junk. Its debt-to-enterprise value sits far above the hyperscaler average. JPMorgan and MUFG spent six months placing a $38 billion debt package for Oracle-linked data centers, and reportedly hit friction distributing the paper.
The plumbing is getting clogged.
Independent Evidence #5: HTX Research and Nassim Taleb
Two independent voices, same conclusion.
HTX Research, the research arm of crypto exchange HTX, published a midyear 2026 report calling AI a “liquidity black hole” absorbing global risk capital. Their thesis: from late 2024 to mid-2026, newly created dollar liquidity flowed disproportionately into AI equities, AI credit, AI private funds, and AI data-center loans. Everything else — crypto, commodities, non-tech equities — got starved.
When a crypto research shop is warning that AI is sucking up too much capital, you know the rotation is extreme.
Then there’s Nassim Taleb. The Black Swan author. The man who built a career on structural risk that everyone else ignores.
In February 2026, Taleb told Bloomberg that bankruptcies in the software space are “definite.” He said tail risk across sectors is “structurally underpriced” and that the risk is “not a small correction — it’s a large drawdown.”
“Someone will make a lot of money in AI,” Taleb said. “It’s not guaranteed to be the companies that make up the AI trade today.”
The Steel Man Assembled
Here’s the strongest version of Rickards’ argument — not Rickards’ version, but the version assembled from independent evidence.
The largest bank in America created a tradeable instrument for betting against AI company debt. The bank that underwrites tech IPOs says the supercycle ends if credit markets close. The most famous contrarian investor of this generation shared a scenario where AI causes systemic collapse. A crypto research firm says AI is a liquidity black hole. A Nobel-tier risk theorist says software bankruptcies are definite.
They’re all pointing at the same structural risk. They’re approaching it from different angles — credit derivatives, equity research, scenario analysis, liquidity analysis, tail-risk modeling. But they’re all describing the same thing: a debt-funded infrastructure buildout that has never been stress-tested by rising rates or falling demand.
The ZeroHedge and Mises Institute piece titled “The Real Threat Is Artificial Credit, Not Artificial Intelligence” captured the macro version: AI is a higher-order capital good being financed by historically distorted interest rates. When the monetary signal normalizes, the malinvestment gets exposed.
What the Steel Man Doesn’t Prove
The steel man doesn’t prove timing.
July 29 might pass quietly. The 80% drop Rickards warns about might take three years — or never happen. The hyperscalers have strong balance sheets, massive cash flows, and genuine pricing power. Microsoft, Alphabet, and Meta are not Lehman Brothers.
But the structural risk is real. It’s independently validated. And it’s being priced by the people who price risk for a living.
JPMorgan didn’t build that CDS basket for fun. Morgan Stanley didn’t publish that $2.9 trillion analysis because they were bored. Burry didn’t share that report because he was looking for Twitter likes.
The strongest version of Rickards’ thesis doesn’t need Rickards to be right about the timing. It only needs the structural conditions to be real.
They are.
Related reading: The AI Black Paper — Jim Rickards’ Thesis JPMorgan’s AI Debt CDS Basket — What It Means Michael Burry’s AI Warning — Deconstructing the Signal
Filed by Sarge · Field Manual · ai-debt · steel-man · analysis · jim-rickards · jpmorgan