Artificial intelligence (AI) has been a tremendous growth opportunity for tech companies in recent years. It's led to them investing heavily in data centers, chips, and other crucial infrastructure to ensure they don't fall behind their rivals. While it's still debatable just how much of a payoff there will be from all these investments, it has led to surging valuations for many top tech companies.
But one person who is a bit skeptical on the hype in AI is Michael Burry, hedge fund manager and founder of Scion Capital. He's known for predicting the housing crash in 2008. Now, he's turned his attention to AI, where he believes that companies are overestimating the useful life of chips. While that may not seem like a big deal, it could mean that companies are underestimating their expenses, and that could be a big problem for investors.
Image source: Getty Images.
A longer useful life means lower depreciation costs
Estimating the useful life of an asset is crucial in accounting because that's how long of a period a company will be spreading its cost over. The longer the useful life, the lower cost per year. There can be an incentive there for companies to overestimate useful life for the sake of lowering expenses and boosting profits. Investors might scoff at depreciation expenses since they are further down the income statement and they aren't directly related to a company's current day-to-day operating activities, but they can have a material impact on earnings.
If earnings are overstated that means the valuations of these already expensive stocks are even higher than they appear to be. Chipmaker Nvidia (NVDA +2.82%), the most valuable company in the world, trades at 45 times its trailing earnings, which may seem a bit high. But its forward price-to-earnings multiple, which is based on analyst expectations, is relatively modest at 23. That's only slightly higher than the S&P 500 average of 21. It's a chip supplier, and its sales numbers suggest that demand remains robust, which may indeed suggest frequent refresh cycles.
In Nvidia's most recent earnings, the company reported $57 billion in sales for the period ending Oct. 26, which was a year-over-year increase of 62%. CEO Jensen Huang says that "Blackwell sales are off the charts, and cloud GPUs are sold out."
Companies may need to upgrade sooner than investors expect
If Burry's claims are true, then it may not be too long before a troubling pattern, which is a frequent upgrade cycle. If chips are only useful for two to three years, as Burry suggests they might be (versus the five-plus years that hyperscalers are claiming), then that could mean more frequent refresh cycles. Companies may be ramping up capital expenditures every couple of years to stay relevant in what's proving to be an intense AI race.
While this might be good news for Nvidia as it could suggest continually high investments into its chips, it could also put pressure on hyperscalers to pull back on some spending, especially if AI projects aren't paying off but investments remain high.
I believe it's a plausible scenario that chips may be useful for only a few years and that lifespan estimates may be a bit high. With companies rushing out to get ahead of one another, it seems unlikely that a hyperscaler is going to be willing to sit on five- or six-year-old chips. The AI boom could put pressure on hyperscalers to invest more heavily in chips and AI infrastructure on a more frequent basis than investors expect.
NASDAQ: NVDA
Key Data Points
Should you be worried about an AI bubble?
If Burry's prediction turns out to be true and companies are setting longer useful lives for AI infrastructure that's unrealistic, then it could indeed be problematic for the tech sector. There's already growing concern that businesses are investing too heavily in AI. If those costs prove to be even higher, then that will only exacerbate those worries.
A bubble may have already formed, and stocks with inflated valuations may be among the most vulnerable to a correction in the near future. With Nvidia trading at a fairly reasonable multiple, it may not be in as bad a shape as other businesses. But the problem is that if there's any sign of softness in AI-related demand, all stocks related to the hype could be due for significant declines, including Nvidia.
Now may be a good time to consider your overall exposure to AI stocks and your risk tolerance. If you're investing for the long haul and still have many investing years to go, then sticking with a blue-chip stock like Nvidia can remain a good move. But if you want to reduce some of your exposure to tech, you may want to consider moving some of your money into exchange-traded funds or indexes that track the S&P 500.
