History doesn't repeat, but it does rhyme, and right now the AI boom is humming the same tune as the dot-com era—only the instruments are bigger and the stakes higher.
The Summary
- AI investment now accounts for roughly a third of US economic growth, creating a scale dependency that makes any correction potentially recession-triggering
- Five key gauges tracking AI bubble risk currently show demand justifying the spending—but the same gauges highlight growing lock-in that makes unwinding harder
- The veteran analyst who lived through the dot-com crash says most money being invested in AI right now will probably be lost, even as AI transforms everything
- The consensus: both sides are right. This is simultaneously a rational buildout of transformative infrastructure and a speculative frenzy that will leave bodies
The Signal
The pattern is familiar to anyone who remembers 1999. New technology arrives. Smart money pours in. Dumb money follows. Prices detach from fundamentals. Then gravity reasserts itself. The early AI boom will likely end with a gigantic bust that clobbers first-wave companies and investors, followed by a long boom that builds the actual winners. Canals, trains, electricity, the internet—they all followed this script.
But there's a critical difference this time. The dot-com era was venture-scale froth. AI spending is macro-scale foundation. When AI investment represents a third of US GDP growth, you're not talking about a sector correction. You're talking about an economy-wide event that triggers bear markets and major recession.
"I don't run an AI company or trade frequently, but I do own stocks and benefit from a healthy economy. So if there's an AI bust, I'll get poleaxed along with everyone else."
The Exponential View analysis tracks five specific gauges for bubble risk: valuation multiples, revenue growth sustainability, capital efficiency, competitive moat durability, and exit liquidity. Right now, none are flashing red. Demand is real. Companies are seeing returns. The spending is being absorbed.
But—and this is the trap door—the spending is also getting harder to unwind. Enterprises are baking AI into core workflows. Hyperscalers are signing decade-long power contracts for data centers. Chip fabs are being built with three-year lead times. This isn't software you can depreciate and move on from. This is infrastructure with sunk costs measured in hundreds of billions.
Key lock-in factors creating systemic fragility:
- Long-term contracts for compute and power that can't be canceled without massive penalties
- Organizational restructuring around AI-native workflows that can't easily revert
- Talent acquisition at bubble-era compensation that creates fixed costs
- Political commitments to AI leadership that make retreat look like failure
The dot-com bust hurt VCs and day traders. An AI bust would hurt millions of people because the exposure is distributed across pension funds, index investors, and the real economy. When a third of growth comes from one category, the whole system becomes fragile to that category's reversal.
The Implication
If you're building in AI, plan for the bust even if it never comes. Keep burn low. Focus on revenue, not valuation. The companies that survive the shakeout are the ones with actual customers paying actual money for measurable value. Don't optimize for the next funding round. Optimize for being the last one standing when capital gets scarce.
If you're investing, know what you own. Is this a picks-and-shovels infrastructure play that wins regardless of which apps succeed? Or is this a consumer app that needs the bubble to keep inflating? The former might survive a correction. The latter probably won't. And if you're just riding index funds, understand that you're now heavily exposed to AI whether you chose to be or not. A third of growth means a third of risk.