Is the AI boom a bubble waiting to pop? Here’s what history says

The answer isn’t so simple, at least according to history.

The S&P 500 Index jumped 16% in 2025, with AI winners Nvidia Corp., Alphabet Inc., Broadcom Inc. and Microsoft Corp. contributing the most. But at the same time, concerns are mounting about the hundreds of billions of dollars Big Tech has pledged to spend on AI infrastructure. Capital expenditures from Microsoft, Alphabet, Amazon.com Inc. and Meta Platforms Inc. are expected to rise 34% to roughly $440 billion combined over the next year, according to data compiled by Bloomberg. 

Meanwhile, OpenAI has committed to spending more than $1 trillion on AI infrastructure, an eye-popping number for a closely held company that isn’t profitable. But perhaps even more troubling is the circular nature of many of its arrangements, in which investments and spending go back and forth between OpenAI and a few publicly traded tech giants.

Throughout history, over-investment has been a common theme when there’s a technological advancement that will transform society, according to Invesco chief global market strategist Brian Levitt, who pointed to the development of railroads, electricity and the internet. This time may be no different.

“At some point the infrastructure build may exceed what the economy will need over a short period of time,” he said. “But that doesn’t mean that the rail tracks weren’t finished or the internet didn’t become a thing, right?”

Still, with equity valuations creeping up and the S&P 500 just posting its third straight year of double-digit percentage gains, it makes sense that investors are growing concerned about how much upside is left and how much market value could be lost if AI doesn’t live up to the hype. Nvidia, Microsoft, Alphabet, Amazon.com, Broadcom and Meta Platforms account for almost 30% of the S&P 500, so an AI selloff would hit the index hard.  

“A bubble likely crashes on a bear market,” said Gene Goldman, chief investment officer at Cetera Financial Group, who doesn’t believe AI stocks are in a bubble. “We just don’t see a bear market anytime soon.” 

Here’s how today’s AI boom stacks up against previous market bubbles. 

Pace, Length

One simple way of gaging whether the AI-fueled tech rally has gone too far or too fast is to compare it against past bull runs. Looking at 10 equity bubbles from around the world since 1900, they lasted just over two-and-a-half years on average with a trough-to-peak gain of 244%, according to research by Bank of America strategist Michael Hartnett.

By comparison, the AI-driven rally is in its third year, with the S&P 500 rising 79% since the end of 2022 and the tech-heavy Nasdaq 100 Index gaining 130%. 

While it’s difficult to draw any conclusions from the data, Hartnett warns investors against fleeing the stock market even if they believe it’s in a bubble because the last stretch of the rally is typically the steepest, and missing out would be costly. One way to hedge is to buy cheap value plays like UK stocks and energy companies, he said.

Concentration

The S&P 500’s 10 biggest stocks now account for roughly 40% of the index, a level of concentration not seen since the 1960s. That has put some investors off, including Wall Street research veteran Ed Yardeni, who said in December that it no longer makes sense to recommend overweighting tech stocks.

Market historians argue that, while the concentration seems extreme relative to recent memory, there are precedents. Top stocks as a share of the US market were at similar levels in the 1930s and 1960s, according to London Business School professor Paul Marsh, who studied the past 125 years of global asset returns. In 1900, 63% of US market value was tied to railroad stocks, compared with 37% tied to technology at the end of 2024, Marsh said.

Fundamentals

Asset bubbles tend to be much harder to spot in real time than after the fact because fundamentals are usually at the center of the debate, and the metrics investors focus on can be fluid, according to TS Lombard economist Dario Perkins. 

“It is easy for tech enthusiasts to claim that ‘it’s different now’ and that fundamental valuations will never be the same again,” he said.

But some fundamentals are always important. For example, compared with the dot-com bubble, today’s AI giants have lower debt-to-earnings ratios than, say, WorldCom Inc. And companies like Nvidia and Meta Platforms are already reporting strong profit growth from AI, which wasn’t necessarily the case in the speculative era 25 years ago.

The potential for credit risk in the AI trade is making some investors nervous. After Oracle Corp. sold $18 billion in bonds on Sept. 24, the stock plunged 5.6% the next day and it’s down 37% since then. Meta, Alphabet and Oracle will need to raise $86 billion combined in 2026 alone, according to an estimate by Societe Generale. 

Valuations

The S&P 500’s valuation is the highest it’s ever been except for the early 2000s, at least according to its cyclically adjusted price-to-earnings ratio, a metric invented by economist Robert Shiller that divides a stock price by the average of its inflation-adjusted earnings over the past 10 years. 

Bullish investors argue that while market valuations are rising because of tech, the pace of increase is much slower than the dot-com era. At one point in 2000, Cisco Systems Inc. was priced at over 200 times its previous 12 months of earnings, while Nvidia is at less than 50 times today. 

Stock prices decouple from earnings growth in an environment where there’s no debate on valuations, according to Richard Clode, a fund manager at Janus Henderson. “We’re just not seeing that currently as yet,” he said.

Investor Scrutiny

Discussions of a potential stock bubble percolated throughout the year but picked up significantly in November and December amid warnings from investor Michael Burry and the Bank of England. More than 12,000 stories in November mentioned the phrase “AI bubble,” roughly equal to the prior ten months combined, according to data compiled by Bloomberg.

Investors see an AI bubble as the biggest “tail risk” event, a December poll by Bank of America showed. More than half of the respondents said the Magnificent Seven tech stocks were Wall Street’s most crowded trade.

This contrasts with the dot-com bubble, when there was “complete excitement about the internet revolutionizing everything,” said Venu Krishna, head of US equity strategy at Barclays. And the questions about whether AI investments will pay off are increasing as the debt issuance rises.

“I wouldn’t brush it off, but I would generally think that scrutiny is healthy,” he said. “In fact, that scrutiny is what will prevent extreme moves like a crash.”