The Next AI Bubble Will Not Look Like the Last One

Every few months, someone warns that artificial intelligence is the next dot com bubble. They may be right about the bubble. They may be wrong about the ending.

Many assume that if AI is overvalued today, it must eventually collapse in the same spectacular fashion as internet stocks did in 2000. That assumption overlooks one important fact. Markets learn.

The dot com crash did not simply disappear into history. It became embedded in the collective memory of investors, venture capitalists, regulators, boards of directors, and entrepreneurs. The painful lessons of excessive valuations, unrealistic business models, and indiscriminate speculation have shaped how capital is deployed today.

This is what I call market memory.

Market memory is more than investors remembering the past. It is the process by which past crises become embedded into the financial system itself. Investment committees become more disciplined. Venture capital firms stage their funding rounds differently. Public markets demand clearer paths to profitability. Regulators pay closer attention to systemic risks. Even founders learn to present their businesses differently because they know what previous generations experienced.

As a result, the next bubble rarely bursts in exactly the same way as the last one.

That does not mean bubbles disappear. Quite the opposite. Human optimism, greed, fear, and innovation remain constant. Capital will always chase transformative technologies. AI is undoubtedly one of those technologies.

What changes is the path of the correction.

Instead of a single dramatic collapse that sweeps away an entire industry, we are more likely to see something more fragmented. Weak companies built on hype rather than substance may fail. Businesses without sustainable economics may quietly disappear. Valuations may compress over several years instead of several months. Capital may gradually migrate from speculative ventures towards companies creating genuine long term value.

The correction becomes evolutionary rather than catastrophic.

Ironically, this may make the AI bubble more difficult to recognise. Dramatic crashes are obvious. Slow reallocations of capital are not. Investors may mistake gradual deterioration for stability, even as weaker firms slowly lose access to funding.

This idea reflects a broader principle that I explore in my economics trilogy, Almost Successful: Macroeconomics at a Threshold, Fragile by Design: Why Modern Economies Remember, Break, and Refuse to Reset, and Before the Choice: Microeconomics in a World That Remembers.

The central argument is that economies possess memory.

Every financial crisis leaves behind more than losses. It changes behaviour. It reshapes institutions. It influences regulation. It alters incentives. These changes become part of the economic system itself, affecting how future booms and busts unfold.

Economies, therefore, are not machines that repeatedly cycle through identical patterns. They are adaptive systems. Every crisis leaves an imprint that changes the probabilities of the next one.

History is often said to repeat itself. I believe a more accurate statement is that history evolves because the system remembers.

If that is true, then the question is no longer whether AI will experience a correction.

The more interesting question is this.

How will market memory change the shape of the next one?