AI Buildout Echoes 1873 Railroad Mania, Historian Warns

The rush to build artificial-intelligence infrastructure is starting to look a lot like the railroad mania that ended in the Panic of 1873, according to Liaquat Ahamed, the Pulitzer-winning historian of that crash.

Ahamed drew a direct line between the two eras in a recent interview. In the 1850s and 1860s, railroad bond issuance swelled to 5% of U.S. gross domestic product as bankers like Jay Cooke poured money into transcontinental lines. By 1873, competition had crushed railroad profitability, the cost of capital was rising, and Cooke could not raise the money to finish his second railroad. His failure triggered a global panic that wiped out half of America's 500 railroad companies within five years.

Today, Ahamed said, "all of the hyperscalers trying to build AI infrastructure at once" face a similar dynamic. The price of AI tokens is already collapsing, he noted, and the cost of capital is ticking up. The big technology companies have begun borrowing to fund their buildouts after relying on internal cash flows. Ahamed said the equivalent of Cooke's collapse would be an OpenAI announcement that it had miscalculated and needed to sell to Microsoft – a scenario that would panic the AI infrastructure market.

That risk is not hypothetical. Data-center builders across the U.S. are racing to sell majority equity stakes worth tens of billions of dollars this summer, according to people familiar with the efforts. The sale process is running into fierce local opposition. Ravi Purohit, co-head of infrastructure at Paul Weiss, told the publication that NIMBYism is being "downplayed" and that buyers will scrutinize how developers handle community relations. "The more they can demonstrate to buyers that they have a constructive relationship with these communities…that actually goes a long way," he said.

Microsoft is one of the hyperscalers at the center of the AI buildout. Its Alpha Score sits at 69 out of 100, a Moderate rating, with the stock at $464.72, up 3% on the session. Ahamed's historical lens suggests the combination of falling token prices and rising capital costs could eventually squeeze even the largest players.

Ahamed pointed to a second parallel: the way the 1873 crash followed a false period of calm. After the Vienna stock market crashed in May 1873, markets stabilized for three or four months. Investors told themselves it was a local incident. Then Cooke's failure hit and the contagion spread globally. "Everyone said, 'If Jay Cooke, a friend of the president and the most well-connected banker in the United States, can't raise $100 million, what hope do we have?'" Ahamed recalled.

The lesson for today, he said, is that the AI infrastructure boom has been rational up to a point – just as the railroad boom was before the overinvestment became obvious. The question is whether the hyperscalers will recognize the turning point before the market does it for them.