Amazon after the dot-com crash and what it means for AI bubble survivors

What does the record say about Amazon after the dot-com crash?

The record is brief. Wikipedia lists Amazon among the companies that gained market share and came to dominate their fields after the bubble, and it adds that Amazon lost a large portion of its market capitalisation during the crash (Wikipedia, Dot-com bubble). Both statements are true together, which is the whole lesson: a company can be hit hard on price and still win on business.

This page keeps to that verified core and leaves out share prices, revenue figures, and dates that it could not check against a primary source. Where a number would help, the page tells you which public document holds it, so you can read the figure yourself and date it correctly.

Why can Amazon’s price fall while its customers keep buying?

Price reflects what buyers expect to earn from owning shares, while use reflects what customers get from the product. In a bubble, expectations run ahead of use. When expectations reset, the price can drop by a large share even if every customer keeps ordering. The two lines move apart, and the gap marks the size of the bubble.

For an AI company, the same split gives you two questions. Do customers use the product more each quarter? And what price does the market place on that use? A rising first line with a collapsing second line describes a survivor in a crash. A falling first line describes a firm that only funding was holding up.

Which public documents let you check an Amazon-style survivor?

Listed companies file annual and quarterly reports with their securities regulator, and the reports hold revenue, costs, cash, and debt. Shareholder letters add management’s view of customers and spending. Read three items for any company: revenue growth, cash left, and the cost of serving one more customer. Together they test the survivor traits without any opinion in the way.

Private AI companies publish less, so ask for the same three items in a data room or contract review. A supplier that shares these numbers lets you judge staying power. A supplier that keeps them hidden leaves you to guess. The choice between those two situations is itself a signal, and it costs nothing to ask.

Which AI businesses resemble Amazon after the dot-com crash?

Look for businesses whose customers pay for a result they use every week, and whose cost of serving each customer falls as volume grows. These businesses can lose a large part of their market value in a correction and still keep their customers. The structure matches the Amazon record better than the size of the valuation does.

Look also at funding. A firm that covers its running costs from sales can wait out a correction. A firm that covers them from new share sales cannot. The EQOS Asia AI study records about $200B a year invested in AI (EQOS Asia AI study, April 2026), so the share of that spending that comes back as customer payments is the number to find.

How would a scoring example work for an AI survivor?

Take an illustrative AI company, invented for this page. Score it from 0 to 3 on each of 4 traits. Customers pay: 3, because contracts renew. Costs fall: 2, because the cost per answer is dropping. Cash lasts: 1, because only 14 months of cash remain. Daily use: 3, because users log in every week. The total is 9 out of 12.

A score of 9 signals promise with one weak point, the cash. The next step is a plan for that point, such as slower hiring or a funding round before the runway shortens. A scoring example does the same job as a medical test: it points to where to look, and it leaves the judgement to the reader.

What does the AI bubble add to the Amazon comparison?

The study records AI valuations up ×4 in 18 months and 95% of AI projects failing in production (MIT 2025, cited by the study). It scores the enterprise AI market at 0.337, in the fragile zone. A fragile market widens the gap between survivors and others, because a shock tips the weakest first.

Amazon’s record gives hope to companies with paying customers. The study gives a warning to companies without them. Both can be true in one market, which is why a bubble produces survivors and failures at the same time. Sorting the two groups early is the practical value of the comparison.

What should you do with the Amazon lesson in an AI bubble?

Pick one AI supplier or investment, collect the three items from the documents above, and score the 4 traits. Repeat the score each quarter and keep the notes. The EQOS Asia method page shows how scores combine into a measurement that anyone can recompute. This page is general information, not investment advice.

The Λ simulator returns a φ(S) estimate in 10 minutes, and a 30-minute pre-diagnostic applies the scoring to a decision in Singapore, ASEAN, or Hong Kong. You bring the decision, and the session returns a score and a ranked set of options.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile