AI Bubble Warning Signs: What to Monitor Without Predicting a Crash

A useful AI bubble watchlist tracks changes in evidence. It does not attach a crash date to a historical chart. The purpose is to identify which assumptions are becoming less plausible and how exposed a particular company is to those assumptions.

Watch customer economics#

Start with paid usage, renewals and the customer outcomes disclosed by the business. User growth and promotional activity are not interchangeable with profitable demand.

Look for explanations when growth depends on unusually generous incentives. Incentives can be a normal customer-acquisition expense; their presence alone does not establish weak demand. The question is whether customers stay when terms normalize.

Watch funding and cash conversion#

Compare collections with reported revenue, and examine debt maturities and investment commitments. BIS Bulletin 120 provides dated research context on the financing of the AI boom. It does not give a trading signal for any stock.

A single increase in receivables may reflect growth or timing. Repeated deterioration, coupled with weaker customer disclosures and tighter funding, deserves a different level of attention.

Separate three scenarios#

In a valuation reset, a company can keep growing while the market pays a lower multiple. In an operating slowdown, sales or margins weaken. In a financing shock, the business struggles to fund obligations regardless of its long-run potential.

These scenarios can reinforce one another, but they imply different evidence and different consequences. Avoid presenting them as one inevitable event.

Record evidence that would change your mind#

For a bearish thesis, specify what would count against it: sustained customer cash generation, reliable renewals or improving returns on investment. For a bullish thesis, specify what would weaken it.

This makes the review less vulnerable to interpreting every headline as confirmation. Revisit the assumptions after substantive disclosures, not every price tick.

No indicator here establishes that an AI crash is imminent. The objective is a clearer view of conditional risks, not a confident prediction that the available evidence cannot support.

More in this analysis series#

Start with AI stocks versus railroad mania. Continue with AI Can Change the World and Still Disappoint Investors.

Analysis published October 1, 2026. Historical comparisons and hypothetical examples are not forecasts. Educational information, not investment advice.