Nobody can predict exactly where AI is headed. But we've lived through technology booms before, and some of them rhyme with today. Three moments are worth revisiting: two that everyone talks about, and one that I think business leaders are overlooking.
1. The Dot-Com Bubble (1995–2001): The Comparison Everyone Makes
The dot-com comparison is both relevant and beaten to death. From Netscape's IPO in 1995 to the NASDAQ's peak in March 2000, investors poured money into anything with ".com" in its name. When it broke, the NASDAQ fell about 78% from its peak, and companies like Pets.com and Webvan disappeared.
If you want a sobering look at how exposed today's market is to AI, Patrick Boyle's breakdown is worth your time.
The optimistic counterargument is that today's leaders are real businesses. Anthropic, for example, has reported adjusted profitability ahead of its expected IPO, and some see that as evidence of a soft landing. I'd ask two questions: how much work is "adjusted" doing in that sentence, and isn't that exactly the number you'd want to announce before going public? (Railway investors in the 1840s learned that lesson the hard way. More on that below.)
It's also worth remembering that the dot-com crash didn't kill the internet. Amazon lost over 90% of its stock value and survived. The technology was real; the valuations weren't.
2. The Luddites (1811–1816): The Backlash Nobody Expects
The Luddites were skilled textile workers watching a well-paid trade disappear as mill owners brought in wide weaving frames and cheaper, less skilled labor. They organized under "General Ludd," a probably fictional leader whose name was signed on threatening letters to factory owners, and smashed the machines that threatened their livelihoods.
The response was severe. Parliament made machine-breaking a capital offense in 1812, and the government sent thousands of troops to the industrial north. Lord Byron gave his first speech in the House of Lords defending the workers. As Brian Merchant argues in Blood in the Machine (2023), the Luddites weren't against technology itself. They were against technology being used to cut their pay and erase their bargaining power.
It's tempting to dismiss this as too extreme for the modern world. But backlash doesn't need sledgehammers. It shows up as regulation, labor action, consumer distrust, and, increasingly, as AI infrastructure becoming a strategic target in conflicts like the war between Russia and Ukraine. The idea of resistance to AI is less remote than it looks.
3. Railway Mania (1845–1847): The Danger Is the Believers
This is the one I think deserves the most attention, because Railway Mania wasn't caused by skeptics or by workers resisting change. It was caused by the adopters.
Early railways were a spectacular success. Connecting a hard-to-reach town to the network unlocked commerce that was already there, waiting: farms, mills, and mines that suddenly had a way to move goods. The rail created value because there was freight waiting for it.
Investors concluded that this success was infinitely repeatable. In 1846 alone, Parliament authorized more than 270 railway acts covering roughly 9,500 miles of new line. Lines were proposed to tiny villages, and competing companies built duplicate routes between the same towns. But many of those villages had no industry to come alive. No freight was waiting.
The mania pulled in some of the smartest people of the era. Charles Darwin lost money on railway shares. The Brontë sisters invested in them. George Hudson, the "Railway King," controlled a huge share of the network until 1849, when it emerged he had been paying dividends out of new capital rather than actual earnings. When the bubble burst in 1847, a large share of the authorized mileage was never built.
Here's the part that matters most: the railways were real, and they did change the world. Much of the network built in that era still carries trains in Britain today. The technology wasn't the mistake. The mistake was believing it would create value in every place it was applied.
What this means for AI in your business
That's the trap I see in a lot of AI strategy right now. Smart people who correctly see that AI is transformative are concluding that it will be transformative everywhere. An MIT study in 2025 found that the large majority of enterprise generative AI pilots were producing no measurable impact on the bottom line. That's a lot of track laid to villages with no freight.
Before you lay track, ask:
- Is there freight waiting? Is there an existing, high-volume workflow with real cost or friction, or are you hoping AI creates demand that doesn't exist yet?
- Are you building a duplicate line? Does this project solve something your current tools already handle well enough?
- Who runs the station? Is there an owner who will maintain, measure, and improve the system after launch?
- What are your real numbers? Are you measuring outcomes, or Hudson-style "adjusted" results that look good in a slide deck?
The rhyme
The dot-com bubble shows what happens when valuations outrun the business. The Luddites show what happens when the people affected by a technology are left out of its gains. Railway Mania shows something quieter and maybe more relevant to most of us: what happens when true believers apply a world-changing technology everywhere except where it actually works.
AI will be worth building. The question for your business isn't whether to lay track. It's where the freight is.