Tech bubbles are as old as technological progress itself. A powerful new technology appears, capital rushes in, expectations race ahead of reality, and prices detach from fundamentals. Then comes the correction. Investors lose money. Yet years later, the technology often reshapes the economy.
This pattern has repeated for nearly two centuries—and it is playing out again with artificial intelligence.
The classic examples
Railway mania (1840s Britain)
Steam trains promised to transform the country. Investors funded far more track than traffic could support. When the boom collapsed, many companies failed. The surviving network became essential infrastructure.
The 1920s boom
Radio, automobiles, and electricity excited markets. Optimism and leverage amplified both the rise and the eventual crash. The technologies endured and became everyday life.
The dot-com bubble (1995–2000)
The internet was correctly seen as revolutionary. Valuations soared, fiber networks were overbuilt, and the Nasdaq fell nearly 80%. Hundreds of companies vanished. The infrastructure that survived later powered search, e-commerce, cloud computing, and social media—creating far more value than was destroyed.
The AI bubble today
The current AI wave shows many of the same traits. A handful of companies—Nvidia, Microsoft, Google, Amazon, Meta, and a few others have driven a large share of market gains on the promise of generative AI and large language models. Capital expenditure on chips, data centers, and power infrastructure has reached extraordinary levels. Secondary startups and related stocks trade at high multiples relative to current profits. The narrative is powerful: AI will change everything.
There are clear bubble-like features—concentrated leadership, heavy front-loaded investment, and optimism that often runs ahead of proven, scaled returns. The biggest players remain highly profitable companies with strong balance sheets, which shapes how any correction is likely to unfold.
What the future may hold
History suggests two outcomes can happen together. A meaningful correction is possible—and in some form, probable. Returns on the massive investments may disappoint in the near term. Competition will intensify. Many secondary AI companies and applications will fail or be acquired cheaply. Data-center and power capacity could temporarily exceed demand, just as fiber did after the dot-com crash. Stock prices in the sector could underperform for years.
Yet the core technology is real. Generative AI and related advances are already improving productivity in software, research, customer service, and content creation. Overbuilt infrastructure, if it occurs, tends to become cheap later and accelerates the next phase of adoption. As with railways, electricity, and the internet, the technology itself is likely to deliver large long-term economic value even if many early speculative bets do not.
The enduring lesson
Tech bubbles are inefficient but effective. They accelerate deployment of talent and capacity. They destroy speculative capital. And they almost always leave behind useful infrastructure and knowledge that the next generation builds upon.
The AI boom fits the historical pattern. A sharp market correction and capital reallocation would be normal. The technology is likely to endure. Many of the early fortunes may not.
