But one very big detail was left unattended: the companies were unable to deliver a durable product that would last through a hot summer. Years more work was needed before a patented process would lead to reliably stable wares, with a big payoff in the U.S. with Civil War demand for rainproof clothes for soldiers.

Another example occurred from 1996 to 2005. Advances in fiber-optic technology that vastly boosted capacity “created a business boom that became a bubble that burst with losses in the trillion dollar range,” said science and technology writer Jeff Hecht, a CalTech engineering grad, on website internet history.org. At that time, the common wisdom was that demand for internet service would double every month. “Dazzled by the new technology,” Hecht wrote, “investors were throwing money at anything optical and innovative.” Frothy valuations were the norm.

Historical comparisons do have their limits. But it is important to pay attention not just to what’s possible technologically with AI but also to the matching of demand and supply.

The Wall Street Journal noted on Sept. 25 that tech companies are pouring hundreds of millions into data centers, most of it borrowed, while AI revenue remains “tiny.” Amazon, Microsoft, Alphabet and Meta alone plan to spend $40 billion in capital investments. Data center leasing firm Coreweave has $42 billion of contracts, $15 billion in debt and is “on the hook” for $56 billion in lease payments over the next decade. Wall Street firms are investing in purchases of parking lots to site data centers, the New York Times reports. One Wall Street Journal columnist suggested a stress test for investments, citing the possibility of a “data center overbuild.”

Industry design and construction firms are already making money on data centers and others plan to make more. What happens if the air goes out of the AI and data center boom? Payments may slow or stop, and those down the payment chain face the most peril. A healthy dash of prudence is wise.