Every transformative technology of the past 150 years has created a bubble -- from railroads to radio to the internet. As much as investors today hope it will be "different this time," it won't.
So as the S&P 500 reaches historic highs, the question isn't whether artificial intelligence (AI) will produce a bubble, but where in the cycle we are: Is the AI bubble about to burst, or is it just beginning to inflate?
Who is using AI and how much?
In many ways, the adoption of modern AI has been remarkable. Just over three years after ChatGPT kicked off the current boom, OpenAI reports that more than 800 million people use its large language models (LLMs) weekly. And as of November 2025, 41% of American workers reported using AI for their work, according to research from the Federal Reserve Bank of St. Louis -- up 10% from a year earlier.
Adoption is widespread at the company level, especially in the information sector, where 37.5% of companies actively use AI, according to research by The Motley Fool.
But these numbers paint a rosier picture than the reality on the ground. While many workers have now tried the technology, only 13% report using it every day. On average, Americans spend 5.7% of their working hours with AI. While that's meaningful, it's not the sweeping transformation the valuations imply -- at least yet.
The bull case: Signs the AI boom is just getting started
There is certainly a compelling case that we are just getting started. Bulls will rightly point out that capital expenditures from the biggest tech companies in the world are still accelerating -- Meta, Microsoft, Alphabet, and Amazon have signaled north of $500 billion in combined AI-related capital expenditures for this year. Companies don't spend that kind of money without seeing a major opportunity.
And while the parallels to the dot-com bubble are hard to miss, there is a key distinction: The companies at the heart of today's boom are enormously profitable businesses with massive cash flows. These aren't Pets.com.
Critically, the technology is still improving. We are entering a phase where so-called agentic AI -- systems that can autonomously execute multistep tasks -- is nearing prime time. If these systems mature to the point where they can reliably handle the kinds of complex workflows that currently require a human, the economic implications would be enormous. That's a big if, but the trajectory of improvement is real, and dismissing it outright would be a mistake.
The bear case: Signs the AI boom is about to bust
But there are serious reasons for caution. First, stocks -- especially those tied to AI -- are trading at extreme multiples. The cyclically adjusted price-to-earnings ratio, a measure of the overall valuation of the stock market, is far above normal levels. In fact, it has been higher only twice: at the dot-com peak and briefly when the global economy ground to a halt in the wake of Covid-19 and earnings collapsed.
Then there's the circularity problem. A significant portion of the AI revenue driving these extreme valuations is being generated by companies selling to other companies. Relatively little of this revenue is coming from end users of AI -- and that organic, external revenue is the linchpin of the whole system. This must change dramatically for the math to work out.
Hidden risks: The debt factor
There's a critical variable that doesn't get enough attention: debt. Leverage has played a central role in every major bubble of the past, and this cycle is no exception. There is an enormous amount of debt embedded in the AI ecosystem, particularly in the build-out of the data centers that power it. Companies like CoreWeave have taken on staggering levels of leverage to finance their infrastructure, betting that AI demand will grow fast enough to service that debt.
When capital is cheap and freely available -- as it was in the lead-up to 1929, 2000, and 2008 -- the system hums along without a hitch. But when rates spike or lending tightens, overleveraged players are the first to crack. There are several macro forces that could trigger this sooner than many expect: sticky inflation, a sluggish labor market, geopolitical tensions, and escalating global conflicts.
Image source: Getty Images.
The bottom line
These massive investments need to generate real returns from outside the AI ecosystem, and I believe investors are starting to lose patience. They are getting increasingly nervous that the return on investment won't materialize.
While adoption rates are impressive, the actual value delivered remains unclear. A recent Bank of England survey found that nine out of 10 senior managers said their company's AI initiatives had produced no measurable impact on labor productivity. That's a staggering disconnect from the trillions of dollars in market value these initiatives are supposedly justifying.
What should investors do?
I believe we are much closer to a major correction than we are to the early innings. I think we could see things materially change within the next year or so.
But I want to be clear: I don't have a crystal ball, and timing the market doesn't work. I am not saying you should sell everything in a panic -- far from it.
Staying invested over the long term has been the winning formula throughout the modern stock market, so I think the smarter move is to use this moment as a stress test.
Take a hard look at your portfolio and ask yourself if you believe in the companies within it. Are they companies that can survive a major drawdown? Will they thrive on the other side?
