AI’s Future May Be Written in Railroads’ Past

The historian A.J.P. Taylor once quipped that the only thing people learn from the mistakes of the past is how to make new mistakes. But there are nevertheless many lessons to be learned if you look in the right places. Consider the extraordinary parallels between today’s AI boom and the 19th-century railway boom, parallels that are pregnant with both warnings and lessons.

The similarities between the railway and digital revolutions have always been striking. The railway revolution shrank physical space. The digital revolution opened and then organized cyber space. The railway revolution transformed every bit of the economy that it touched. Ditto the digital revolution. Both revolutions were masterminded by a handful of titans who became mega-rich in the process.

Yet the parallels have multiplied since the arrival of AI. Hitherto, the digital empires have been asset lite while the railways were necessarily asset heavy. AI is forcing IT companies to spend as never before in the physical economy, with McKinsey and Co. predicting that they will invest $5.2 trillion on chips and data centers in the next five years.

US capital spending on AI currently accounts for 1.5% of the US economy. That is roughly the same as US capital spending on the railroads in the 1860s. The “Magnificent Seven” technology firms account for more than one-third of the S&P 500. The railroad companies accounted for about 60% of the US stock market in the 1880s. OpenAI, a company that was only founded in 2015, is valued at $500 billion, more than any European company; it expects to burn through $115 billion of cash by 2029.

The first lesson is that building ahead of demand all but guarantees booms and busts. Building rail lines in the middle of nowhere requires a big upfront investment. But once you have built the railway, civilization floods in and the value of the wilderness rises exponentially. The promise of outside returns created periodic manias: From 1868 to 1871, US rail companies built 33,000 miles of tracks. It also encouraged a buccaneering business style: relentless boosterism, extravagant financial engineering and competitive over-building. Charles Francis Adams, a railroad reformer whom Congress forced the Union Pacific to hire as its president from 1884 to 1890, argued that his industry’s “method of doing business is founded upon lying, cheating and stealing.”

Booms inevitably led to busts. The great financial panics of 19th-century America, in 1837, 1857, 1873 and 1893, were all triggered by the railroads. In 1873, for example, a group of railway men borrowed heavily from Jay Cooke & Company, a major component of the country’s banking establishment, to build a second trans-continental railway only to find themselves unable to pay back the loan. Cooke & Co. collapsed; other banks followed; the stock market plunged; and the “long depression” of 1873-90 followed.

The second lesson is that giant egos can make all this instability worse. The 19th-century railways were a strange hybrid: To run properly, the railways demanded order, with hierarchies of managers, precise timetables and detailed accounts, and yet they were summoned into life by buccaneers who, for the most part, did not give a fig for the rules of conventional business. What mattered to them was grabbing the biggest prize and humiliating their rivals.

US entrepreneurs have seized the limelight in this buccaneers’ tale: the spider-like Jay Gould watering stocks and the cunning Leland Stanford manipulating politicians. But Django Davidson, of Hosking Partners, argues that it is time to turn the spotlight on a Brit. George Hudson was so synonymous with the 1840s railway boom that he was known as “the Railway King.” By the mid-1840s, the modestly born Hudson controlled 1,000 miles of railway lines, about a quarter of the country’s stock, owned the biggest private house in the country, on the edge of Hyde Park, and got himself invited to dine at Buckingham Palace. He attracted capital by his genius for publicity — persuading everybody (including Queen Victoria who traveled on a special train that he provided) that railways were the big new thing and that anybody who missed out was a fool. “He took away people’s breath at first,” said the Times obituary, “but he soon succeeded in persuading them that the larger the project and the bolder the scheme, the more likely it was to pay.” He was eventually forced to flee his creditors and hide out in Paris.

The similarities between then and now are striking, right down to the invitation to Sam Altman, the poster boy of AI, to dine with King Charles III during Donald Trump’s official visit. The digital companies are investing heavily well before revenue streams come online on the grounds that AI eventually will become indispensable. Some of the biggest digital companies such as Meta Platforms Inc. and Oracle Corp. have taken the unusual step of issuing bonds. The digital companies are also doing the equivalent of building a second continental railroad by building competing AI machines.

The trendsetters in the AI world talk about artificial general intelligence, or smarter-than-human super intelligence, in the same mystical tones that the railway barons once talked about manifest destiny. They are paying astonishing sums to hire the people who might be able to discover this modern Snark, with Mark Zuckerberg offering $200-million pay packages to poach AI superstars from rivals and startups. And conventional managers who try to introduce some discipline into this frenzy are pushed aside, as happened with the directors of OpenAI who briefly ousted Sam Altman.

The overall lesson for investors and policymakers alike is beware. The Magnificent Seven are certainly much better capitalized than any of the 19th-century railroads. And the stock markets are both much better run and much more diversified. Yet some complex deal structures, such as OpenAI’s financing through vendor agreements, resemble the circular arrangements of past bubbles. And the sheer volumes of investment in data centers and chips are not justified by consumers’ willingness to part with their money. Richard White, the author of Railroaded: The Transcontinentals and the Making of Modern America (2011) points out that the railroad boom was bad for regular investors, who frequently lost their shirts; bad for the financial system, which was frequently shaken to the core; but good for the business people who presided over the system and accumulated some of the world’s biggest fortunes.

There was one other great beneficiary of the railway boom, however: the public in general, particularly future generations. The boosterism may have set off many wild goose chases. The booms and busts may have ruined lives and shaken markets. But the economic pluses were huge: the shrinking of time and space, the reduction in the cost of economic inputs, and higher productivity. Whether the benefits of AI will be equally large awaits to be seen, but it is clearly making all sorts of tasks, from organizing our affairs to searching for information, significantly easier. Discussions of AI often question whether it is a speculative bubble or a transformative technology. The biggest lesson of the railroad mania of the 19th century is that you can be both, big time.