Tulips, railways, dot-com: where the AI bubble fits in bubble history

Why does financial bubble history help you judge the AI bubble?

History gives you cases where the outcome is known. Each past bubble shows how prices, funding, and stories behaved before the fall, and the AI bubble can be tested against those behaviours. The comparison cannot predict a date, and it does supply questions that you can ask of today’s numbers, which is the practical benefit.

Use history as a checklist, not as a template. Every bubble had a different asset, a different country, and a different funding source. Matching the features, and not the surface story, keeps the comparison honest. A case that matches 4 of 5 features deserves more weight than a case that matches one.

Which five features does every financial bubble share?

The first feature is a new idea that changes how people see value. The second is easy money, through credit or low rates. The third is a story that explains why old valuation rules no longer apply. The fourth is a wave of new buyers who join because prices rose. The fifth is a funding gap: spending that revenue has not yet repaid.

Score each case from 0 to 5 on these features. For AI, the EQOS Asia AI study records valuations up ×4 in 18 months and about $200B a year invested, which speaks to the story and the funding gap (EQOS Asia AI study, April 2026). Easy money and the new buyers need current data on credit and on retail flows, so update those two each quarter.

What does tulip mania add to the history of the AI bubble?

Tulip mania in the Netherlands peaked in the winter of 1636 and 1637, and prices fell abruptly in February 1637. Much of the trading used contracts for bulbs, not the bulbs themselves, which shows how a market can trade a promise. The page on tulip mania vs the AI bubble explains what the comparison gets right and where it stops.

The tulip case fits the pure story and new buyer features. It does not fit the AI case on productive use, since a bulb produces little beyond a flower, while AI tools have users and revenue. For the AI bubble the lesson is narrow: a price can rest on the next buyer, so check who the next buyer is.

What do the railway booms teach about the AI bubble?

Railway mania in Britain ran through the 1840s and funded real infrastructure. Investors could buy shares with a deposit of 10%, which let a small sum control a large position, and the build-out continued after prices fell. The page on railway mania vs the AI bubble compares infrastructure booms and asks who owns the assets afterwards.

This case is the closest to AI on the physical side. Data centres, chips, and power lines resemble tracks and stations: useful assets that investors may pay too much for. The lesson is that a bubble can destroy investor wealth and still leave valuable infrastructure. Both can be true, and the question is who bears the loss.

What does Japan’s 1989 bubble teach investors about the AI bubble?

Japan’s stock index peaked in December 1989, after a boom in shares and land. The decades that followed showed that a bubble can end in a long period of slow growth and not in a quick recovery. The page on Japan’s 1989 bubble vs the AI bubble draws a lesson for Asian investors who hold concentrated positions.

This case adds a time dimension. Recovery took far longer than most investors expected, so the horizon of your savings matters as much as the size of the fall. For readers in Singapore, ASEAN, and Hong Kong, it also shows how a local market can follow a global theme. Check your own horizon before you read any forecast.

Where does the AI bubble sit on the historical scorecard?

Dot-com in 2000 matches the story and the new buyers. The mother guide, AI bubble vs dot-com bubble, compares that period in detail. Tulips match the story. Railways match the physical build-out and the funding gap. Japan matches the long aftermath. AI matches parts of all four, and no case matches every feature.

A mixed score is informative. It says that the AI bubble question cannot be settled by naming one analogy, and that each analogy points to a different risk: a fall in price, a loss on assets, or a long slowdown. Track the features each quarter and watch which analogy fits better as new data arrives.

How does EQOS Asia turn bubble history into a decision?

The EQOS Asia AI study scores the enterprise AI market at φ(S) = 0.337, in the fragile zone where a shock can tip the system (EQOS Asia AI study, April 2026). The EQOS Asia method page explains how the score is built, so you can set the number beside your historical scorecard.

A 30-minute pre-diagnostic applies the scoring to your own case in Singapore, ASEAN, or Hong Kong. You bring the decision, such as an AI budget or a portfolio position, and the session returns a score and a ranked set of options. History informs the questions, and the score gives you an answer you can recompute.

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