Ever since ChatGPT exploded into the mainstream in late 2022, the question has followed AI everywhere it goes. Is this an AI bubble?

By mid-2023, headlines were already warning about it. By 2024 and 2025, the idea had hardened into something close to consensus. Analysts, researchers, skeptics, and even people building AI systems started saying the same thing. Something feels overheated.

But calling something a bubble is easy. Understanding why it might be one is harder.

Speculation alone does not explain what is happening. Massive investment alone does not either. Even the fact that companies like Nvidia briefly rivaled entire national economies in market value does not automatically mean collapse. In theory, the world could decide AI is worth it all.

The real question is this. What kind of bubble are we dealing with, if it is one at all?

To answer that, it helps to stop guessing and look at how tech bubbles actually form.

How Tech Bubbles Usually Work

Economists Brent Goldfarb and David Kirsch studied dozens of major technologies across history, from electric lighting to aviation to the dotcom era. Their conclusion was not that bubbles are random or irrational accidents. They form in predictable ways.

According to their research, bubbles tend to appear when four conditions show up together.

High uncertainty about what the technology will become
Companies built entirely around that technology
New or inexperienced investors rushing in
A powerful story that makes the future feel inevitable

When those four line up, markets stop asking careful questions and start chasing narratives.

That framework gives us a useful way to look at the AI bubble conversation without panic or denial.

Uncertainty Is Everywhere in AI

Some technologies are immediately easy to understand. Electricity made sense the moment people saw lights turn on. Cars replaced horses in a way that was obvious.

AI is not like that.

Nearly three years into the boom, there is still no clear answer to what the long-term business model for most AI companies actually is. Are these tools replacing search engines? Automating office work? Becoming digital assistants? Something else entirely?

Costs remain extremely high. Many AI companies lose money on each user interaction. Energy use is expensive. Computing infrastructure is limited. Legal questions around training data and copyright are unresolved. A major MIT study found that the vast majority of companies using generative AI have not made money from it.

Even the goals keep shifting. One moment the focus is artificial general intelligence. The next, that term is dismissed as vague. Then it is replaced by talk of superintelligence.

When uncertainty does not shrink over time, it grows into fuel for bubbles.

Pure-Play Companies Raise the Stakes

Another warning sign is the rise of companies whose entire future depends on AI succeeding exactly as promised.

These are called pure plays. They do not just use the technology. They are the bet.

Nvidia is the most obvious example, having tied its growth almost entirely to AI demand. Others include companies like OpenAI, CoreWeave, and Perplexity. Investors are not just betting on steady progress. They are betting on a specific version of the future arriving fast.

What makes this riskier is how interconnected these companies have become. Chip makers rely on AI labs. AI labs rely on cloud providers. Cloud providers rely on AI demand to justify massive infrastructure spending.

When one link weakens, pressure spreads quickly.

Retail Investors Are Now Deeply Involved

Unlike earlier eras, this boom is not limited to insiders.

Retail investors are pouring money into AI-related stocks through trading apps and index funds. Nvidia alone attracted tens of billions of dollars from individual investors in a single year.

This matters because when expectations outrun reality, everyday people feel the consequences. Retirement funds, savings accounts, and pensions become exposed to the same story-driven risks once reserved for venture capitalists.

Everyone is a novice when the technology is this new. That makes enthusiasm easy to spread and hard to slow down.

The Power of the Story

Every major tech bubble has had a moment when belief snapped into alignment.

For aviation, it was the first transatlantic flight. For radio, it was the realization that voices could reach millions at once. These moments told investors the future had arrived, even if the business fundamentals were not ready.

AI has had several of these moments already. The release of ChatGPT was one. The race for larger models was another. The promise is always framed the same way.

AI will transform everything. Jobs, medicine, education, climate science, national security. If you hesitate, you fall behind. If you regulate, you lose.

It is an incredibly strong story. Stronger than most technologies ever had. And unlike earlier examples, the promise of AI feels limitless rather than specific.

That is what makes the AI bubble feel different.

So Is This Really an AI Bubble?

By the historical definition, it checks every box. There is deep uncertainty about what AI will ultimately become, a growing number of high-profile pure-play companies whose futures depend entirely on it, widespread participation from inexperienced investors, and a powerful narrative that frames AI’s rise as inevitable.

That does not mean AI is useless or fake. Many technologies that later reshaped the world passed through bubbles first. Radio survived its crash. Aviation did too. The internet followed the same pattern.

The problem is timing. Markets often collapse long before a technology has the chance to mature.

The real danger is not that AI disappears. It is that expectations fall faster than reality can possibly catch up.

If history is any guide, the question is not whether the AI bubble bursts, but who gets caught standing too close when it does.

Buyer beware.