The Paradox of Historical Bubbles

According to Bloomberg Opinion columnist Justin Fox, financial history is littered with manias that destroyed capital, yet it offers almost no reliable signal for when a bubble will burst. The challenge facing investors today is not whether AI represents real technological progress—it likely does—but whether the investment frenzy surrounding it has created structural vulnerabilities that could trigger a painful correction.

Fox's framing captures a central tension in bubble analysis: transformative technologies and speculative excess are not mutually exclusive. The electric light, the automobile, and the railroad were genuine innovations. The existence of real value does not prevent bad capital allocation or investor losses.

Debt and Sentiment as Leading Risk Factors

Rather than attempting to pinpoint a crash date, Fox identifies two structural vulnerabilities that historical precedent suggests matter: the level of debt embedded in the system, and the state of investor and creditor sentiment. These are not forecasting tools but risk indicators—they describe the brittleness of the market structure, not its timing.

Debt is a critical variable because it creates forced sellers. When an investor buys on margin or when a company finances expansion through leveraged borrowing, a price decline creates immediate pressure to liquidate. The Wall Street Crash of 1929 was devastating not merely because stock prices fell, but because the pervasive use of margin lending meant declines triggered forced selling, which in turn triggered more margin calls, which triggered more selling—a cascade that turned a price correction into a 89% decline over three years. Margin lending in 1929 represented roughly $8.5 billion, equal to the entire currency in circulation at the time.

Sentiment determines whether creditors and investors are willing to absorb losses or whether they panic. In the Mississippi Company Bubble of 1719–1720, John Law's system depended entirely on confidence that the banknotes it issued would retain value. The moment large holders—notably the Prince de Conti—began converting notes to gold and attempting to exit, confidence collapsed. The state's response (making it illegal to own more than 500 livres in coin) destroyed what little confidence remained. By 1721, the Company's shares were worthless and the banknotes were repudiated.

When Technology Meets Circular Finance

The AI sector's current structure exhibits both real technological potential and genuine financial risk. Major technology companies and infrastructure providers are deploying billions in capital expenditure for data centres, chips, and software. This is not speculative—it represents real productive investment. However, the financing mechanism that has enabled this buildout is partly circular: AI-focused companies have issued debt at valuations dependent on expectations of future profitability; investors have poured capital into AI infrastructure equities and debt instruments; and AI companies have used their rising equity valuations as collateral or justification for further leverage.

This circularity is not stable if three conditions reverse: (1) if revenue growth disappoints relative to capital spending, (2) if credit becomes scarcer or more expensive, or (3) if sentiment shifts from assuming continued growth to questioning whether returns justify the capital deployed. Any one of these could force a repricing; all three together could force widespread liquidation.

Bubblepedia Analysis: Structural Parallels to Historical Episodes

Concentration of capital in unproven business models. In the dot-com era (1995–2000), capital flowed disproportionately to companies with minimal revenue and no clear path to profitability. The comparison between the dot-com bubble and AI infrastructure shows that while the underlying technologies (the internet and AI) were real, investor willingness to fund businesses with no demonstrated returns was not.

Leverage without revenue validation. The Wall Street Crash emerged from a decade of leveraged speculation where margin debt reached unsustainable levels before underlying earnings growth faltered.

Sentiment as the break point.

What History Cannot Tell Us

Fox's key insight is that historical precedent warns of risk but does not forecast timing. We know that high leverage and exuberant sentiment create fragility. We do not know whether that fragility will trigger a 20% correction, a 50% decline, or a catastrophic crash—nor do we know whether it will occur in months or years. The 1920s boom lasted a decade; the Mississippi mania lasted less than a year. Market psychology and credit availability operate on their own timescales, indifferent to technological fundamentals.

This ambiguity is frustrating for investors seeking certainty, but it reflects reality. No historical precedent can tell us at what equity valuation, debt level, or interest rate the AI sector becomes "too expensive." These thresholds depend on participant expectations, which are inherently forward-looking and prone to sudden revision. A technology can remain valuable while investors suffer severe losses, if they overpaid at peak mania and must revalue at more sustainable multiples.

The Risk Monitor: Debt Growth and Credit Conditions

Rather than predicting a crash, Fox's framework suggests watching three signals: (1) the rate of debt accumulation in AI-adjacent sectors (semiconductors, data centres, software infrastructure), (2) the terms on which new debt can be issued (tightening terms or rising refinancing costs signal sentiment shift), and (3) the behaviour of large holders when market conditions weaken. The first two are measurable; the third requires real-time judgement.

When debt growth accelerates faster than revenue growth, when refinancing becomes more difficult or expensive, or when institutional holders begin to liquidate rather than buy dips, the system has begun to shift from expansion to contraction. At that point, the speed of repricing depends on how much leverage is embedded and how quickly creditors respond. History suggests these reversals, once underway, can accelerate sharply.

Conclusion: Real Technology, Uncertain Timing, Measurable Risk

AI is likely a transformative technology. The companies deploying it at scale may generate substantial returns for patient investors. But the investment frenzy, the leverage in the system, and the speed at which sentiment can shift mean that losses for many participants—even in a genuinely valuable sector—are not speculative but historical. The question is not whether the technology justifies capital spending, but whether investors have already overpaid for the rights to future returns. History suggests that answer is often yes, and that the market's correction of that error can be violent.

Further reading and context

South Sea Bubble (1720).

The Everything Bubble (2009-2022), The Global Bond Bubble and Zero Interest Rate Era (2009-2022).