Is There an AI Bubble? The Nifty Fifty Show It Isn’t That Simple

Nvidia has become the undisputed leader in a stock market powered by AI hopes. Valuations are lofty for both the U.S. chip maker and the technology giants that are buying its products to train their AI models—a group dubbed the Magnificent Seven. Some investors worry a bubble may be inflating.

The famous “Nifty Fifty" stocks of the 1970s are one guide to today’s hopes and fears. They were also powerful blue-chip corporations with extravagant valuations. Whether they ended up justifying them is still a matter of debate.

In a famous 1998 article, Wharton School professor Jeremy Siegel argued that they did. A portfolio containing all of them had a lower total return than U.S. stocks overall, he showed, but by a minuscule margin and only relative to the very peak of the market in December 1972.

Four years later, Pomona College’s Jeff Fesenmaier and Gary Smith disputed this. Since there was never an official list of the Nifty Fifty, Siegel included many stocks such as Philip Morris and General Electric, which had price-to-earnings ratios under 25. Defined as the 50 priciest stocks, the Nifty Fifty would have severely underperformed.

Some stocks are common to both lists—a group that Fesenmaier and Smith dubbed the Terrific 24. Had investors split $1,000 equally between them in December 1972, they would now have $95,000 in the bank, following a particularly dismal performance since the pandemic. If investors had instead bought the S&P Composite 1500 index, they would have $185,000. Only four firms in the list—McDonald’s, Eli Lilly, Merck and Texas Instruments—have outperformed the market during this period.

Yet Siegel still had a point. In the 51 years since December 1972, only four of the Terrific 24 have gone bust. Two of those—Polaroid and Eastman Kodak—were caused by the same event: the decline of analog photography. Of the 12 that remain stand-alone businesses, all still count as big blue chips, with the arguable exceptions of Xerox and IFF.

Sure, most of 1972’s hot stocks face cooler expectations in today’s market, which favors “new economy" firms. But the resilience of these big companies is still impressive.

Contrast that with the late 1990s dot-com rally, which featured myriad loss-making online-economy ventures. A few thrived, notably Amazon.com, eBay and Booking.com. Some survived in obscurity, such as Ask Jeeves, which was acquired by IAC. Among the 70-odd internet stocks that rocketed in 1999, though, most were completely wiped out.

Or take the blank-check mania of 2021. In some cases, the revenue-light startups that went public—air taxi manufacturers, for example—wanted to apply tech that didn’t yet exist to a market that didn’t yet exist. Solactive’s De-Spac Index, which tracks former blank-check vehicles, is down 90% from its peak.

Of course, not overvaluing the earnings potential of established businesses is important. Even within the Terrific 24, picking the cheapest five stocks led to far better results: $1,000 would have turned into $248,000.

But this is different from speculating whether there is a business at all, be it in hydrogen trucks today, online sales of pet food in the early 2000s, or monopoly profits for the British South Sea Company in the 18th century. This is the true meaning of a bubble: If it pops, you lose it all.

The AI rally can appear bubbly because the magnitude of the commercial application of large language models remains anyone’s guess. Venture Capital is now flocking to a panoply of startups, most of which will doubtless fail.

But it is hard to see a true bubble in the stock market. For one, initial public offerings remain subdued. Also, a business case for improving user interfaces through generative AI is almost certain to exist for blue-chip software companies like Alphabet, which are already investing massively in chips. Nvidia’s sales more than tripled in the fourth quarter of 2023 relative to a year earlier. This is why the stock’s price-to-earnings ratio is 31, below the recent average despite a 240% gain over one year.

The case of Cisco illustrates where the real risk lies. While its price-to-earnings ratio of 126 in the year 2000 was clearly overblown, a big part of the problem ended up being the volatility of the earnings. The issue wasn’t being in a bubble, but selling networking equipment to firms that were.

When the digital rush ended, leaving the market oversupplied, Cisco’s profit growth suddenly halved. Investors have since come to value it as an unexciting hardware firm, and its shares are still shy of their dot-com peak.

This is the scenario that should trouble Nvidia and other chip companies, such as AMD, Arm and ASML. In their favor, they have a far more financially reliable customer base than Cisco did in 2000.

When it comes to blue-chip stocks, the term bubble is a misnomer, even if vast profits are being spent and made to prepare for a business with highly uncertain prospects. What should such a market be called instead? Answers on a postcard.

Write to Jon Sindreu at jon.sindreu@wsj.com