The story you have heard about the tulip bubble is not quite the story that happened. The corrected version is more interesting and slightly less morally satisfying. It also bears on something else.

What did happen began in the late winter of 1635 in the brick taverns of Haarlem and Alkmaar, the great florist towns north of Amsterdam, and ended on Saturday 3 February 1637 at a Haarlem auction at which no one bid. Between those two dates, a network of perhaps several hundred Dutch professional florists, wealthy merchants, and skilled craftsmen traded futures contracts on tulip bulbs that were, in many cases, still in the ground. The rarest varieties, hybrids whose striped or feathered petals were the product of a virus that the Dutch did not understand and could not reliably reproduce, changed hands at prices that climbed through the autumn of 1636 and the first weeks of 1637 until, in the first week of February, the bidding stopped. By summer the value of outstanding contracts had collapsed to about three and a half per cent of nominal. The Court of Holland declined to enforce the contracts. The province's florists' guilds met in May and agreed to settle on those reduced terms. A handful of merchants went bankrupt. There are no recorded suicides. The Dutch economy continued its expansion. The East India Company paid its annual dividend on schedule. The Eighty Years' War with Spain continued to be financed without disruption.

That is what happened.

The story that has been told for almost four hundred years is different. In the older telling, all of Dutch society went mad for tulip bulbs. Cobblers and chimney sweeps mortgaged their houses to buy them. A sailor mistakenly ate a rare bulb for breakfast and was nearly arrested. A Semper Augustus, the rarest variety, traded for the price of an Amsterdam canal house plus a coach and four. When the bubble burst, the entire Dutch economy collapsed, fortunes were ruined, men killed themselves, the country was permanently traumatised, and the lesson that human nature is incurably foolish was written on the wall.

Almost none of this is true.

The principal historian who has reconstructed what actually happened, Anne Goldgar of King's College London, spent a decade in the Dutch civic archives identifying the names of every traceable tulip buyer and seller of 1635-37. She found that the trade was concentrated in a network of perhaps three hundred wealthy merchants and skilled craftsmen, almost all of whom belonged to the same religious congregations and many of whom were related by marriage. She found fewer than half a dozen documented bankruptcies in which tulips can be identified as the primary cause. She found no suicides. The infamous sailor story comes from a moralising pamphlet of 1637 and cannot be corroborated. The famous Semper Augustus contract is from the same pamphlet and is probably notional. The "all of Dutch society" framing is contradicted by the Court of Holland's records, which show that the trade was a private matter within an identifiable social group and was settled within that group's institutions.

How did the false version travel? Through pamphlets. In the spring of 1637 a Haarlem printer named Adriaen Roman published a satirical dialogue, Samen-spraek tusschen Waermondt ende Gaergoedt, between an honest Calvinist citizen named Truth-Mouth and a foolish speculator named Greedy-Greed, in which Greedy-Greed loses his fortune on tulips and is morally instructed by Truth-Mouth. Other pamphlets followed. Engravings appeared: most famously the Floraes Mallewagen, Flora's Cart of Fools, in which the goddess of flowers is depicted leading a wagonload of bankrupt speculators into the sea. These were not financial documents. They were religious-ethical interventions, written by Calvinist printers in the context of a long-running Dutch civic argument about the moral hazards of speculation, luxury, novelty, and the new mercantile wealth that the Republic's Iberian war was generating in such uncomfortable quantities. They sold well. They were reprinted. They became the primary source for almost every subsequent telling of the tulip episode, including Johann Beckmann's 1797 History of Inventions, and including the most influential telling of all: the three-page chapter on "The Tulipomania" in Charles Mackay's 1841 Extraordinary Popular Delusions and the Madness of Crowds. Mackay was a Scottish journalist writing for a Victorian audience newly anxious about railway speculation. He had no access to the Dutch archives. He summarised Beckmann, who had summarised the pamphlets. The Mackay chapter has been reprinted, paraphrased, summarised and analogised in financial journalism continuously for one hundred and eighty-four years. It is the source of almost every later metaphor.

This is where today comes in.

In the present capital-expenditure cycle for artificial intelligence, the phrase "tulip mania" has appeared in financial-press coverage with a frequency that nobody in the present generation of analysts has previously seen applied to a single technology. The Bank of England warned in October 2025 that US equity valuations, particularly in AI-focused firms, looked "stretched" on metrics "comparable to the peak of the dotcom bubble." It repeated the warning in December 2025, citing the debt-financed nature of the AI infrastructure buildout. The International Monetary Fund's managing director Kristalina Georgieva drew the dot-com comparison publicly in October 2025 and again at Davos in January 2026. Goldman Sachs chief executive David Solomon told an audience this month that "a bunch of the capital being deployed in AI will actually not produce any returns." Former SEC chair Gary Gensler has noted that AI infrastructure capex reached approximately four hundred billion dollars in 2025 and is projected to reach somewhere between five hundred and six hundred billion in 2026, against generative-AI revenues of perhaps fifty billion. OpenAI is preparing a confidential draft IPO prospectus with Goldman Sachs and Morgan Stanley at a private-market valuation reported above eight hundred billion. Anthropic is in talks at a nine-hundred-billion valuation. None of these numbers requires a metaphor to be alarming or reassuring on its own terms. But the metaphor that gets reached for, almost reflexively, is the tulip.

Notice what reaching for it does. The tulip frame imports a moral structure: it tells us that the participants are foolish, that the asset is intrinsically worthless, that the eventual collapse will be a deserved correction, and that the historical record demonstrates a recurring human propensity to mania. None of those four propositions is an empirical claim about AI capex in 2026. Each is a moral inference borrowed from a religious-ethical pamphlet of 1637. They may be correct about AI. They may not be. The point is that they are not what the historical record of the tulip episode actually shows. The historical record shows a narrow professional market in a perishable luxury good, a six-week speculative episode, a private settlement among the traders involved, and a national economy that absorbed the loss without disruption while continuing to finance a war.

What is striking about the 2026 chatter is that the moral frame is doing most of the analytical work. JPMorgan's chief executive Jamie Dimon told an Anthropic event this month that the trillion-dollar AI capex wave is "fundamentally justified" rather than speculative excess. The Bank of England and the IMF have warned. Goldman Sachs has hedged. The Federal Reserve has watched. None of them has the data to settle the question. What everyone has is a metaphor, and the metaphor is older than the data, and the metaphor was written by Calvinist pamphleteers in 1637 to make a religious-ethical point about luxury, and translated to a Victorian railway audience in 1841 by a journalist who had not been to Holland.

This does not tell us whether AI capex will produce the returns its champions promise. That is a question about chip economics, data-centre power demand, model-capability scaling, enterprise adoption rates, and the cost of compute. None of those questions is answered by a tulip. The lesson of the Dutch civic archives is more disquieting than the lesson of the pamphlets, because it is harder to use: it is that when an entire professional class reaches for the same historical analogy, the analogy may be telling us more about the moral preferences of the analogists than about the economics of the asset.

The historians have known this for decades. The pamphleteers are still winning.