Investors' Defensive Moves Show the AI Boom Is a Bubble Likely to Burst
Source: Branko Marcetic. "Wall Street Is Getting Very Nervous About AI." October 23, 2026. jacobin.com
The Gist
The author argues that the people with real money at stake are now acting as if AI is a bubble: they're dumping AI-related debt and buying insurance against defaults. Business numbers, such as huge losses and revenue targets without precedent, back up their worry. Because AI is propping up the economy and is financed with lots of debt, a crash could hurt far more than tech companies.
Conclusion
The AI industry is an economic bubble built on unsustainable finances, and investors' growing defensive behavior shows the bubble is starting to deflate, with serious risks to the wider economy and financial system.
Premises
- Investors, bond traders and insurers are acting on doubts about AI: data center and neocloud bond prices have dropped, CoreWeave's bonds are distressed, nearly half of this year's data center bonds are rated junk, and trading in AI-linked credit default swaps more than quadrupled between July and September.
- The AI build-out's economics are implausible: Brookings estimates about $3.5 trillion in annual revenue (nearly 9% of GDP) will be needed by 2032 to justify the spending, with no historical precedent, and Apollo's chief economist notes that expected tech-sector cash flows far exceed what customer sectors are expected to have.
- Flagship AI firms are deeply unprofitable and fragile, as shown by Character.ai's bankruptcy filing, Anthropic's prospectus showing an $80 billion-plus loss, and OpenAI's delayed IPO and reduced valuation.
- AI firms face major external and structural risks: public opposition has blocked over a hundred data centers and led over five hundred localities to restrict them, and AI chips are becoming obsolete faster than the loan periods that are secured against them.
- Debt financing is large and growing, and lenders are lending on sponsors' ability to absorb liabilities rather than on project economics, which resembles the subprime mortgage crisis according to Carlyle's Jason Thomas.
- AI is a leading pillar of US GDP growth, and the Bank of England and the IMF have flagged its economics as a threat to financial stability, so a bust would spill over into the broader economy.
Assumptions
- Investor behavior such as bond selloffs and CDS purchases is a reliable signal of underlying fundamentals rather than short-term sentiment, liquidity or hedging
- Revenue requirements calculated by Brookings and Gartner are accurate and AI revenue will not grow enough to meet them
- Current losses and compute costs at AI firms are structural, not a temporary stage of scaling
- Local opposition to data centers will keep growing and meaningfully constrain the build-out
- The comparison with the dot-com crash and the subprime crisis is apt for judging systemic risk
- Hedging by some investors indicates a coming collapse rather than ordinary risk management in a still-growing sector