Would Buffett Buy the AI Debt Thesis?

Let me ask you a different question than the one you’re expecting.

I’m not asking whether Warren Buffett would buy the stocks Jim Rickards recommends. I don’t know what those are. Neither do you. That’s not the point of this exercise.

The question is: Would Warren Buffett buy Rickards’ risk assessment?

Because that’s what the AI debt thesis really is. It’s not a stock pick. It’s a warning about hidden leverage — trillions of dollars in AI infrastructure debt that nobody’s talking about, sitting in off-balance-sheet structures and opaque financing vehicles.

And on that question, the answer is fascinating.

Here’s why.

Buffett on Leverage: He Saw This Coming

In 2002, Warren Buffett wrote something in Berkshire Hathaway’s annual report that sounds like it was drafted last week.

He called derivatives “financial weapons of mass destruction, carrying dangers that, while now latent, are potentially lethal.”

That’s not a throwaway line. He spent pages explaining why. He warned about off-balance-sheet entities. He warned about counterparty risk. He warned that when you can’t see who owes what to whom, you’re one domino away from a chain reaction.

Sound familiar?

Rickards’ AI debt thesis makes the same structural argument. Massive capital commitments to AI infrastructure — data centers, GPUs, fiber, power grids — financed through complex vehicles that don’t show up on traditional balance sheets. The argument is that when rates stay higher for longer, that debt gets repriced, and the hidden leverage unwinds fast.

Buffett wouldn’t need to be convinced. He’s been making this exact argument for over two decades.

Buffett on Bubbles: He Doesn’t Buy What He Can’t Value

Here’s where it gets personal.

Buffett sat out the dot-com bubble. From 1995 to 1999, the Nasdaq gained 400%. Technology stocks were printing millionaires overnight. Analysts valued companies on “eyeballs” — website traffic, user counts, anything except earnings.

Buffett wouldn’t touch them.

In the 1999 annual report, he wrote: “We have no insights into which participants in the tech field possess a truly durable competitive advantage.”

He got killed for it in the press. Fortune ran a cover story asking “What’s Wrong with Warren Buffett?” Berkshire shares fell while the Nasdaq soared. He looked like a dinosaur.

Then the Nasdaq dropped 78% from its peak.

And Buffett looked like a genius.

The lesson isn’t that Buffett hates technology. It’s that he won’t pay for something he can’t value. If you can’t figure out how a company makes money ten years from now, you shouldn’t own it.

Now apply that to the AI infrastructure companies issuing this debt. How do you value a data center that’s built on a seven-year depreciation schedule, financed with floating-rate debt, servicing a customer base that could build its own capacity tomorrow?

Buffett would look at that and say the same thing he said in 1999: I don’t know enough to own this.

Buffett on Cash: The $167 Billion Question

By the end of 2024, Berkshire Hathaway was sitting on $167.6 billion in cash.

That’s a record. And it’s not because Buffett got scared.

Buffett doesn’t hoard cash out of fear. He hoards it because he doesn’t see enough fairly priced opportunities. The most successful investor in history — the guy who bought Coca-Cola, Geico, and See’s Candy — is looking at the market in 2024-2025 and saying, “I can’t find enough things worth buying at these prices.”

Think about what that means for a market where AI infrastructure companies are issuing debt at aggressive rates to fund buildouts with no visibility on returns.

If Buffett won’t buy Coca-Cola at 30 times earnings, why would he buy a data center REIT with a seven-year window and a floating-rate loan?

Buffett on AI: He’s Not Anti-AI. He’s Anti-Speculation.

Let me be precise about something.

Buffett owns Apple. Apple is arguably the largest AI-adjacent company in the world. As of early 2025, Berkshire still held roughly 300 million shares worth tens of billions of dollars.

He’s not anti-AI. He’s anti-speculation.

The difference? Apple has real cash flow. Real margins. Real products you can hold in your hand. Buffett bought Apple not because it was a “tech stock” but because it was a consumer products company with a moat — a brand so strong that people pay a premium for the logo.

The AI infrastructure companies issuing this debt? They don’t have that.

They have buildouts. They have projections. They have financing vehicles.

They don’t have Coca-Cola’s pricing power or Apple’s ecosystem lock-in.

That’s the distinction Buffett would draw. He’d look at the balance sheets, ask where the recurring revenue is, ask what happens when the buildout cycle ends, and he’d pass.

The Buffett Verdict

Here’s what I think Buffett would actually say about the AI debt thesis — not the stocks, the thesis.

He’d agree with the structural argument.

He’s already made it. The 2002 derivatives warning is the same playbook. Off-balance-sheet exposure, counterparty risk, opacity that masks systemic leverage. Buffett has been warning about this for 20 years. Rickards is basically writing a sequel.

But he’d question the timing.

Buffett doesn’t try to call crashes. He doesn’t short markets. He doesn’t position for catastrophe. He waits for prices to come to him. “Be fearful when others are greedy, greedy when others are fearful” — he wrote that in 1986, and he’s never stopped living by it.

The AI debt thesis could be right about the structure and wrong about the calendar. Buffett would acknowledge the risk and keep holding cash until the opportunity shows up at a price he likes.

What This Teaches You

Here’s the part that matters for every promo you’ll ever read, not just this one.

Buffett’s framework isn’t about predicting crashes. It’s about not paying too much for things you don’t understand.

That’s it. That’s the whole system.

• If you can’t understand how a company makes money, you can’t value it.
• If you can’t value it, you shouldn’t own it.
• If everyone around you is buying it and you’re sitting on cash, you might look stupid for six months. But you won’t lose your money.

This applies to the AI debt thesis. It applies to crypto. It applies to SPACs. It applies to every hype cycle that comes down the pike.

The question isn’t “will the AI debt bubble pop?” It’s “do I know enough to be in this position at this price?”

If the answer is no — and for most of us, on most of these infrastructure plays, it is — then do what Buffett does.

Hold the cash.

Wait for the price that makes sense.

---

Related reading: Steel Man: The AI Debt Thesis — the strongest version of the argument, and AI Black Paper by Jim Rickards — the promo that sparked this analysis.