AI bubbles aren't scary in themselves; what's scary is they're increasingly propped up by debt
Historical data shows equity-driven bubbles cause limited economic damage when they burst, but credit-driven bubbles break bones—and this year mega AI companies' debt issuance jumped from roughly $200 billion to $500 billion.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Stock markets are a side hustle, bond markets are the main business
Wigglesworth compares stock markets to trading Pokémon cards: the vast majority of trades are simply one person selling stock to another, with no impact on the company itself. Bond markets work differently—every day they finance new wastewater plants, roads, factories and other real projects, channeling capital through capitalism's most primitive mechanism. By scale, he notes, the US bond market is three to four times larger than the stock market; globally, it's even bigger than the entire banking system. That's why he calls this book the engine of global economics.
— Robin WigglesworthWartime crisis accidentally invented the modern bond market
In 1171, when Venice and Byzantium went to war, the Doge levied a mandatory «war tax» on citizens proportional to their wealth. To quiet dissent, he promised 5% annual interest. The revolutionary moment came when these receipts could be freely bought and sold. Venice eventually lost that war and never repaid the tax; it became a de facto perpetual bond. But Venetians discovered that tradeable income streams had intrinsic value and could serve as collateral. The model spread: other Italian city-states copied it, then France, Spain, and Holland amplified it further.
— Robin WigglesworthBritain's vast debt burden defeated Napoleon himself
France under Napoleon vastly outmatched Britain militarily, but Britain, with its unified Consols market, could borrow almost unlimited sums and consistently fund allies like Prussia. France's weaker credit meant it paid interest rates two or three times higher. After the war, Britain carried the highest debt-to-GDP ratio in its history. Contemporary doomsayers predicted collapse, but historian Macaulay countered that government bonds could be used as collateral and circulate as money—that debt actually capitalized and strengthened the British economy. This concept is what gives the book its title, «fabulous debt.»
— Robin WigglesworthRailroad bonds built America and sparked a historic depression
After the Civil War, banker Jay Cooke pioneered advertising and circulating bond prospectuses to ordinary citizens, selling US government debt and railroad bonds to the masses. Without bond market finance, transcontinental railroads couldn't have been built. But this debt-fueled infrastructure boom collapsed in 1873: German and Austrian stock bubbles burst, European investors withdrew their US railroad bond holdings, and Cooke's pet project—the Northern Pacific Railroad, which he'd personally underwritten—ran out of money. The ensuing depression, initially called the «Great Recession» until fifty years later the term was replaced with «Long Depression», became the template for financial crises.
— Robin WigglesworthRating agencies aren't as bad as they look
After 2008, rating agencies became synonymous with failure. But when Wigglesworth reviewed crisis retrospectives, he found something surprising: the actual ratings on RMBS and CMBS products were remarkably accurate. Prices collapsed because nobody trusted anything; when final losses were calculated, most ratings weren't far off actual performance. Only some AAA-rated tranches saw unexpected losses. The real problem wasn't grossly missed judgments—it was the fundamental conflict of interest: rating agencies sold ratings to the highest bidder, a structural problem that persists today.
— Robin WigglesworthAI investment may be mismatched again
Wigglesworth was once «profoundly relaxed» about AI bubbles, for two reasons: he doubted his own ability to judge them, and bubbles aren't necessarily bad—infrastructure outlasts them, just as railroads and fiber-optic cables survived their own crashes. His thinking shifted because this wave of AI investment increasingly relies on debt and off-balance-sheet obligations rather than tech giants' internal cash flows. Historically, credit-driven bubbles burst far more violently than equity-driven ones.
— Robin WigglesworthAI debt issuance doubled in a year
The numbers tell the story: mega cloud companies issued roughly $500 billion in debt this year, up from about $200 billion last year. Wigglesworth doesn't think disaster is imminent—these companies generate phenomenal cash flow, Meta, Amazon, YouTube print money continuously, and rating agencies now factor in off-balance-sheet commitments. But he flags Oracle as «relatively weak» in the herd, already showing stress. His read: «We're still at the foot of the mountain», with no guarantee that today's signals won't look like early warnings in hindsight years from now.
— Robin WigglesworthTen years of borrowing equals two centuries
The US accumulated about $20 trillion in debt over its 240-year history, then borrowed another $20 trillion in the past decade alone. Wigglesworth traces this to «a series of self-inflicted errors»—massive tax cuts, discretionary wars, tariffs—not responses to genuine crises. He's skeptical of the standard debt-to-GDP ratio and prefers debt-service costs as a share of GDP: currently around 3.5%, versus 2.5% a few years ago. Historically, countries crack under stress around 5%; the CBO projects the US will hit that threshold in roughly ten to twelve years.
— Robin WigglesworthIn their own words · checked verbatim
If you want to see where capital actually flows, like capitalism, at its most primal, that's the bond market.
Robin Wigglesworth13:27
It was, in fact, a fabulous debt. It was a blessing.
Robin Wigglesworth27:53
They erroneously imagine there was an exact analogy between the case of an individual who was in debt to another individual and the case of a society, which is in debt to a part of itself. That analogy leads to endless mistakes.
Barry Ritholtz28:55
The bond market was the track on which the trains truly ran
Barry Ritholtz36:06
The ratings were, I was staggered to see pretty good.
Robin Wigglesworth54:29
AI hyperscalers, I think, have issued around $ 500 billion worth of debt this year after $ 200 billion or so last year.
Robin Wigglesworth1:05:44
In its first 240 years, the US borrowed around $ 20 trillion. And it's borrowed another $ 20 trillion in just the past 10 years.
Robin Wigglesworth1:07:49
Figures
| US debt accumulated over 240-year history | approximately $20 trillion | 1:07:49 |
| US debt issued in the past decade | approximately $20 trillion | 1:07:49 |
| AI mega-cloud debt issued this year | approximately $500 billion | 1:05:44 |
| AI mega-cloud debt issued last year | approximately $200 billion | 1:05:44 |
| Current US debt-service costs as % of GDP | approximately 3.5% | 1:10:51 |
| US debt-service costs as % of GDP (several years ago) | approximately 2.5% | 1:10:51 |
| Historical debt-service stress threshold (as % of GDP) | approximately 5% | 1:10:51 |
| Louisiana Purchase price | $15 million | 33:04 |
| Current global debt | approximately $370 trillion | 1:12:54 |
Glossary
- Consols
- Perpetual bonds—undated UK government debt once considered the world's safest asset.
- Sukuk
- Islamic bonds that pay no interest but deliver fixed returns through asset-lease structures.
- Securitization
- Bundling large loan portfolios into tradeable bonds and selling them to investors.
- Bearer bond
- Unregistered bonds that any holder can redeem for interest at any agent bank.
- CBO
- US Congressional Budget Office; the independent fiscal forecasting body that models future debt trajectories.
How to listen
Founders, investors, and engineers concerned with US debt sustainability, AI financing structures, and financial history—also works as a story during commutes or workouts.
The opening ten minutes about the author's personal background contributes little to understanding the bond market itself and can be skipped.