Tulip mania vs the AI bubble: what the comparison gets right

What happened during tulip mania, and why is it cited for the AI bubble?

Tulip mania took place in the Dutch Republic. Prices of tulip bulbs rose from 1634 and peaked in the winter of 1636 and 1637, then fell abruptly in February 1637. At the peak, some bulbs sold for more than 10 times a skilled artisan’s yearly income. The story spread because the numbers look absurd and the fall was fast.

Writers cite it for AI because it makes the idea of a bubble easy to grasp: an asset priced far above any use. That simplicity is also the weakness of the comparison. A vivid story can raise alarm without measuring anything, so the useful step is to extract the mechanism and test it against data.

What does tulip mania get right about the AI bubble?

It gets one mechanism right: a price can rest on the next buyer. When buyers pay more because they expect others to pay more later, the price detaches from use. AI shares can show the same pattern when valuations rise ×4 in 18 months while 95% of AI projects fail in production (EQOS Asia AI study, April 2026).

It also shows how a story spreads. Tulips became a topic of conversation among people who had never grown a flower, and new buyers joined because prices rose. Watch for the same signs in AI: investors who enter for the first time, and arguments that depend on a theme more than on a financial statement. A record of these signs, kept with dates, shows whether enthusiasm is spreading or fading, and it gives a board a shared reference for the next discussion.

Where does the tulip mania comparison with the AI bubble stop?

A bulb produced a flower and a future bulb, and little else. AI products produce answers, code, images, and savings, and companies pay for them. The EQOS Asia AI study projects productivity gains of +40% and 300 million jobs that change (EQOS Asia AI study, April 2026). These figures describe real demand that tulips never had.

AI companies also report revenue, profit, and cash flow, so analysts can compare price with earnings. A bulb had no earnings to compare. The right reading is that AI is not a tulip, and a real technology can still carry an overpriced valuation. Price against earnings is the test the tulip story cannot supply. Add revenue per user and the share of customers who renew, and you hold a fuller picture of whether demand is durable or borrowed from future funding rounds.

What do tulip contracts teach about AI spending commitments?

Most trading in the tulip boom used forward contracts for bulbs that were still in the ground, so buyers and sellers traded promises instead of goods. That structure let prices rise without any bulb changing hands. The lesson for AI is to look at commitments: long-term purchase agreements for chips and cloud capacity.

Commitments that exceed expected demand are a risk, because a cancelled promise leaves the seller with a gap. Ask each large AI company how much of its backlog is firm and how much depends on customers who must still raise funds. A firm backlog from paying customers supports prices, while a backlog funded by new investors depends on the next round.

Which four questions turn tulip mania into a test for the AI bubble?

First, does the asset earn cash? Second, who is the marginal buyer: a user, a company, or a speculator? Third, how much trading rests on promises not yet delivered? Fourth, how many new buyers entered in the last year? Answers to these 4 questions place AI nearer to a tulip or nearer to an internet company.

Write the answers with a date and repeat them each quarter. As of October 2026, the first answer favours AI, since many AI products earn cash. The other three remain open and depend on funding data. Open answers explain why a verdict remains a probability, and the probability moves as new numbers arrive.

How does tulip mania fit with the other bubbles in history?

Tulip mania is the pure story case: a price with almost no use behind it. The guide Tulips, railways, dot-com: where the AI bubble fits in bubble history places it beside cases with more real value, such as railways and the internet. Reading all of them together gives a scale from pure story to useful infrastructure.

On that scale AI sits between the poles. It has real users, as railways had, and it has stories of the kind that tulips had. Neither extreme describes it. A mixed position calls for a measured view, with a score that combines evidence, not a slogan borrowed from one case.

How can EQOS Asia help you test a tulip mania claim about AI?

The EQOS Asia AI study scores the enterprise AI market at φ(S) = 0.337, in the fragile zone where the system works and a shock can tip it (EQOS Asia AI study, April 2026). The score gives the funding and structure question a number you can recompute on the Λ simulator.

When a decision depends on the answer, a 30-minute pre-diagnostic applies the scoring to your case in Singapore, ASEAN, or Hong Kong. The session returns a score and a ranked set of options, so you leave with a decision supported by evidence instead of an analogy.

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