Nvidia’s murky AI future isn’t reflected in its price

Every now and then, a company comes along that is so dominant and is growing so quickly that it feels like the only stock anyone cares about. I’m referring, of course, to Nvidia Corp., the chip giant powering artificial intelligence. Its stock has surged 4,000 per cent over the past five years, making it one of the three most valuable businesses in the world alongside heavyweights Microsoft Corp. and Apple Inc.

Nvidia deserves every bit of the attention. Its market share for AI chips is around 90 per cent. Its profit margin is 57 per cent on $80 billion in revenue, by far the highest revenue among companies with comparable profitability in the S&P 500 Index. It has also grown sales by 64 per cent a year over the past five years, the highest growth rate among S&P 500 companies.

Nvidia is clearly a great business, but is it a great investment? To answer that question, one must also consider valuation, often the flipside of profitability. Fast-growing, highly profitable businesses tend to command a premium, and Nvidia is no exception. At 76 times one-year trailing operating earnings, its valuation is more than three times that of the S&P 500 and twice as expensive as Microsoft and Apple.

Investors would undoubtedly argue that they’re betting on Nvidia’s future, not its past, and that future growth justifies its current price. Helpfully, Bloomberg compiles analysts’ “long-term” earnings growth estimates, which represent the annual rate at which analysts expect companies to grow operating earnings per share over their next business cycle, usually three to five years. Nvidia’s consensus long-term growth rate is 43 per cent a year, which is among the highest for S&P 500 companies and multiples that of Microsoft and Apple.

At that rate, Nvidia’s operating earnings per share should climb to $80 in four years from $19 today, resulting in a long-term price/earnings ratio, let’s call it, of 18. Sorting S&P 500 companies from high to low on that measure, Nvidia’s long-term P/E ratio ranks 116 and is nearly identical to that of Microsoft and Apple. In other words, after accounting for Nvidia’s higher expected growth, it’s no more expensive than the other tech titans.

There’s just one catch: The further out analysts try to estimate earnings, the more uncertain their projections become, and that uncertainty is greater for some companies than others. Microsoft and Apple are mature businesses that make money on a captured customer base that includes nearly everyone with a screen. Nvidia, on the other hand, serves a newer, if more promising, market centered around AI, the future of which is less certain, as is Nvidia’s role in it.

One would therefore expect more disagreement about Nvidia’s future growth than that of Microsoft and Apple, and in fact, that is the case. Bloomberg also calculates the standard deviation — or variability — of analysts’ long-term earnings growth estimates. Higher variability signals less agreement among analysts, and vice versa.

It turns out that the variability of Nvidia’s long-term growth estimates is three times that of Microsoft and four times that of Apple. So, while all three have similar long-term P/E ratios, those of Microsoft and Apple are underpinned by more modest but also more stable forecasts, whereas Nvidia’s is based on rosier projections that are harder to pin down.

That highlights the risk of relying on future earnings to justify a stock’s current price, particularly when the outlook is hazy, as it appears to be with Nvidia. If consensus growth forecasts for the company prove to be too optimistic, the stock will reprice, and it has a lot of room to correct.

There have been numerous comparisons of Nvidia to Cisco Systems Inc. recently, and for good reason. Like Nvidia’s chips driving AI today, Cisco built the routers that powered a booming internet in the 1990s, and the frenzy around internet stocks made Cisco the most valuable business in the world for a time.

What is most instructive about that comparison is how misguided Cisco’s investors turned out to be. During the decade from 1991 to 2000, Cisco grew operating earnings by 71 per cent a year, boasting earnings of $4.6 billion for fiscal year 2000 and a hefty valuation of 233 times trailing earnings, no doubt justified by confident forecasts for continued outsize growth.

But when the internet bubble burst in 2000, Cisco’s fortunes turned quickly. Operating earnings plunged to just $21 million in fiscal year 2001, and by the time they recovered their 2000 level two years later, Cisco traded at 39 times trailing earnings — a fraction of its 2000 valuation. Its valuation has trended lower ever since.

That doesn’t mean a similar fate awaits Nvidia, but it does demonstrate the perils of banking on a bright future in a new and fast-moving industry. My Bloomberg Opinion colleague Parmy Olson already sees “signs of discontent” with AI, including businesses cutting back on new AI tools.

“Nvidia’s GPU chips are in essence the new gold or oil in the tech sector,” Daniel Ives, an analyst at Wedbush Securities, wrote recently. Perhaps it’s no coincidence that the path of both commodities is notoriously difficult to predict.