The stock market has broken several records this year, overcoming tariffs and signs of a hard-pressed American consumer, and recovering from a recent slump largely due to the promise of artificial intelligence.
The share prices of some AI companies have risen dramatically in a short time, and technology companies are investing billions to build data centers and microchip factories to drive the boom.
While investors and analysts see good reason to justify the optimism behind the nearly 50% rise in the S&P 500 over the past two years, some warn that current valuations are still based on a big bet on the future.
Continues after advertising
The stock market experienced a similar moment 29 years ago, when Alan Greenspan warned about “excessive irrational optimism” fueling a bubble in internet stocks. The S&P 500 had risen more than 60% in the two years before Greenspan’s speech.
Policymakers are again sounding alarms.
“By some measures, U.S. equity valuations are approaching levels not seen since the dot-com bubble,” noted Andrew Bailey, governor of the Bank of England, in a recent speech.
But when Greenspan made his warning, the dot-com bubble was just beginning to inflate. The S&P 500 would rise more than 100%, that is, double in value, before peaking in March 2000. After that, the index fell for the next two years, cutting its value in half and reaching a bottom at approximately the same level as when Greenspan made his warning six years earlier.
Some investors fear that we are in a period similar to 1999, just before the crash. But most believe we are closer to 1996, and that taking money out of the market now could mean forgoing considerable gains that are yet to come.
“It’s uncertain, but it’s still early days,” said Paul Christopher, head of global investment strategy at Wells Fargo. “This isn’t 1999 — we think we can say that for sure.”
Continues after advertising
AI vs. Dot-com bubble
There are important differences between the current enthusiasm for AI and the dot-com bubble, traders and analysts say.
One way to analyze this is by the price/earnings ratio (P/E) — calculated by dividing the share price by the company’s profit, that is, how much money the company generates. The higher the number, the more speculative the bet.
Continues after advertising
Current OP/L of the S&P 500 is above usual. Over the past 10 years, the index has averaged about 22. Now, that index is around 27, close to 29, which was the peak in 1999, just before the dot-com bubble burst.
“It’s difficult for the stock market to perform well when you get to these valuations,” said James Masserio, head of global equities at RBC Capital Markets.
An important difference is that the companies driving the rise today are also generating more money.
Continues after advertising
In the 1990s and early 2000s, many of the companies leading the stock rally weren’t making much, if any, profit. This led to very high P/E ratios for some companies because stock prices continued to rise even as earnings lagged far behind.
Cisco Systems was the largest company in the S&P 500 when the dot-com bubble burst in March 2000. Its P/E ratio reached over 200, meaning the share price was more than 200 times the previous year’s earnings, and without additional growth, even if all profits were distributed to shareholders, it would take two centuries for the investor to recoup their investment.
A high P/E is, therefore, also a sign that investors are betting that the company will grow substantially, that is, that it will make more profits in the future, and when that happens, the price will appear more reasonable. This means that P/E ratios also show us how much investors today depend on the future turning out as they hope.
Continues after advertising
This never happened for Cisco. The company’s profits grew from 15 cents per share in 1996 to 36 cents in 2000, far below the runaway price appreciation. Today, Cisco is valued at about $300 billion, even less than investors valued it years ago.
Not just blue skies and unicorns
Nvidia, the largest company in the S&P 500 today and the first to reach a market value of $5 trillion, also had moments of extreme valuation. Its P/E ratio reached over 200 in 2023, on par with Cisco, during the initial enthusiasm for AI following the introduction of ChatGPT in November 2022.
But since then, Nvidia has made much more profit. Demand for its high-power chips has yet to be satisfied. Today, its P/E is around 45, and if the calculation is based on expected future earnings rather than the last 12 months, that ratio drops to around 25.
Nvidia’s ability to consistently meet high profit expectations has been “remarkable,” said Alex Altmann, head of tactical equity strategies at Barclays. “It’s almost unbelievable.”
“It’s not all blue skies, rainbows and unicorns, but so far these businesses have demonstrated repeated excellence,” he added, “which needs to be recognized.”
While Nvidia’s share price has risen about 1,000% in the last three years, from $17 to $180, its profits — the real money it’s making — have risen even faster. That means the stock is probably cheaper today than it was three years ago, said Stacy Rasgon, an equity analyst at AB Bernstein.
The weaknesses of AI
Bubbles tend to start with a reasonable thesis before investors’ bets on the future become disconnected from reality. Even when the market falls, the thesis often proves correct. The internet didn’t disappear after the dot-com bubble.
While investors are less concerned about big companies driving the rally than they were 25 years ago, they are concerned about the fervor that has led to unprofitable companies being financed, or companies using debt to invest in AI, as well as the market’s concentration in a few giants.
Those concerns helped spark a sell-off last month that traders say caused investors to exit some top AI stocks and move into other areas of the market.
Oracle lost a third of its value after revealing that it will turn to the debt market to finance its AI expansion, borrowing $18 billion in September. Morningstar analyst Luke Yang predicted that Oracle’s total debt could grow from $100 billion today to $300 billion by 2030 due to the company’s AI plans, highlighting a “very high risk if demand for AI doesn’t materialize as people expect now.”
CoreWeave is down about 42% since October. The new AI cloud computing company relies heavily on large contracts with companies like Nvidia and Microsoft, but it is still unprofitable and has accumulated a large debt.
“There is greater systemic risk due to the intertwined nature of these companies, and they have heavy debt to finance capital expenditures,” Masserio said. “Not all will be winners. Some will be losers, and when they lose, how will that affect other companies?”
New competitive pressures could also emerge that alter the outlook for some of the biggest companies driving the rally. Many investors fear that inflation could accelerate next year, which could lead the Federal Reserve to keep interest rates high. Or there may simply be delays in building the necessary infrastructure and obtaining the energy to power the growing use of AI.
The moment when the bubble may burst is very difficult to predict. Last month’s market slide was reversed by strong earnings and increased confidence that the Fed will continue cutting rates.
Michael Burry, an investor known for predicting the 2008 financial crisis, recently compared Nvidia’s role in the AI craze to Cisco’s during the dot-com boom and described OpenAI as the next Netscape, “doomed and bleeding money.”
c.2025 The New York Times Company
