Don’t go all-in on Nvidia stock. Here’s how to play the AI trade now.

Yes, it has been a tough year for the AI trade. Chip stocks, software stocks, and even power stocks and shares of industrials that build out data centers have tumbled. Much of the decline is captured in the Nasdaq Composite, which is down 13% from its record close in December, while Nvidia, the engine that made the artificial-intelligence trade go, has fallen nearly 30%. Even the S&P 500 utilities sector, home to a handful of companies that provide power for data centers, has dropped 5.8% from its high as they move in the same direction, if not with the same magnitude, of their tech brethren.

The cheapening of AI stocks makes them look more attractive—but caution is warranted. While Nvidia is looking more attractive than it did two months ago, allocating all of one’s AI money to the company comes with the risk that it loses data center chip market share to the likes of Broadcom and Advanced Micro Devices. But buying just those two would mean being exposed to the possibility that Nvidia continues to dominate and they never catch-up to the degree that the market expects.

It doesn’t end there. Buying only Microsoft means accepting a company that analysts expect to see slowing sales growth this year, according to FactSet, which could further reduce its price/earnings multiple. It also means ignoring other, less obvious companies that offer other types of AI software services that companies are just beginning to purchase.

So what’s an investor to do? Buy them all. “Investors must increasingly consider risk management when it comes to stocks with exposure to the important growth themes," writes Trivariate Research’s Adam Parker. “Investing in stocks with the best attributes in each theme makes more sense than just overweighting the best company in any one theme."

The risk management part of the equation is a result of the increasing tendency of AI-exposed stocks to trade in tandem. AI chip stocks—Nvidia, Broadcom, AMD, Micron Technology, Teradyne, Applied Materials, Lam Research and KLA Corporation, among them—have traded with almost their highest correlations historically to utilities in the past 21 days, according to Parker. “[You] are not going to have much diversification from stocks that power Data Centers vs. owning the Semiconductor designers / manufacturers themselves," he writes.

It isn’t just chips. AI software stocks, including Microsoft, Oracle, Salesforce, Adobe, ServiceNow, and Palantir Technologies, have seen close to their highest correlations to utilities in the past 21 days. Industrials that sell electrification products, including Eaton, nVent Electric and Generac, are interesting, too—they have been less correlated to utilities than software and chips have, but have also seen an increase.

So when everything is trading in lockstep, investors should consider buying a basket of stocks in lockstep. The chip stocks, in aggregate, are expected to grow earnings at a double-digit clip annually over the next three years, according to Barron’s calculations of FactSet data. Utilities’ projected growth is in the high single digits because power for data centers doesn’t comprise a majority of their revenue. And if chip stock earnings come in lower than expected because of a sluggish economy, they’d simply grow off that lower base of profits, given that many companies and people worldwide still need to adopt AI. The software and industrial stocks are in a similar boat.

When in doubt, just buy them all, and sleep well at night.

Write to Jacob Sonenshine at jacob.sonenshine@barrons.com