By Jack Hough

Hard-drive stocks are suddenly popular. I don't remember the Bible specifically mentioning that as a sign of the apocalypse, but it's nonetheless jarring.

Wait times for high-capacity drives from Seagate and Western Digital have blown out to nearly a year on ravenous demand from artificial-intelligence customers, according to Benchmark Equity Research. Both companies are raising prices amid epic gains for profits and share prices.

Not so long ago, if data-center components had been ranked by investment sizzle, hard drives would have just edged out cable zip ties. In 2018, when Western Digital was trading at six times projected earnings, I wrote in Barron's that cloud computing demand could lift the stock from "dire" levels to a more traditional discount, rewarding "deep value" investors.

It did no such thing. The centralization of data storage, and technologies that squeezed fuller use out of existing hard drives, left a glut of cheap terabytes.

Now hyperscalers are turning the internet's attic into AI insights, while frantically putting up new data centers that dwarf football stadiums and shopping malls. AI companies have driven pretty much all of the S&P 500's upside since the late-2022 introduction of ChatGPT. Yearly capital investments by hyperscalers are projected to compound at 30% and hit $500 billion by 2027.

So far, newly installed machines are finding plenty of work. It's "super fun seeing people love images in ChatGPT," tweeted OpenAI founder Sam Altman earlier this year, "but our GPUs are melting." Graphics processing units, Nvidia's specialty, are the stars of the AI boom. Altman's near-term fix for his heat problem was to throttle down speeds. Long term, he will need more machines.

This data-center buildout has breathed new life into all sorts of adjacent industries. Earlier this month, a strategist at BofA Securities walked me through why the world will need $3 trillion in fresh spending on nuclear power through 2050 to supply AI with enough watts, and how AI was pulling forward development of next-generation nuclear technologies, like small modular reactors and even fusion, a longtime sci-fi staple.

"For 22 minutes, we had a mini sun on our planet," he said of a recent fusion experiment in France. If nothing else, that seems bullish for tanning.

I've read AI-driven bull cases for everything from air conditioning to earth-moving machines. So why not hard drives, too? Western Digital and Seagate are both up more than 300% over the past three years. Benchmark predicts more upside for each, especially for Seagate, which has a technology edge. Shares there already go for 21 times projected earnings, but apparently deserve to fetch 24 times next year's estimate.

Then again, what if the yearlong wait for hard drives causes customers to double- and triple-order, leading to an inevitable glut, and the next downturn? Or what if the broader AI cashvalanche is a replay of the late 1990s dot-com bubble, when companies borrowed richly to lay "dark fiber, " or broadband capacity that exceeded demand?

It's not just my inner killjoy wondering. There are outer killjoys, too, like hedge fund manager David Einhorn, who this past week -- as AI stocks wobbled -- warned that the companies are spending too much, too quickly. If he's right, the S&P 500 could be vulnerable at 25 times earnings. It's almost enough to make me do something truly nihilistic, like buy a value fund.

Don't sweat the spending, says Barclays Capital. Today's hyperscalers are collectively funding their data-center spree from cash flow, with room to spare for stock buybacks. And there are signs of solid usage beyond Altman's hot cores. Nearly every economic sector has seen a jump in AI-related job listings. Microsoft has reported breakneck growth for sales of AI services. Anthropic, which was founded five years ago by OpenAI defectors and has a higher mix of enterprise versus consumer users, has seen its revenue run rate jump fivefold this year to $5 billion.

Perhaps January's DeepSeek scare provides a more useful comparison than the dot-com bubble. In case you've already forgotten: A Chinese AI company introduced a free chatbot to rival ChatGPT, and Wall Street was startled by how well it performed compared with how little was spent to build it. If a country that lacks access to top chips can Gilligan's Island its way to effective AI, the thinking went, why are U.S. companies spending all this money? Tech and energy stocks were hit hard.

Nvidia plummeted 17% in a day. But an S&P 500 investor lost only 1.5% the same day, and around 70% of index members closed higher, because investors clambered into value, quality, and staples.

Barclays takes the DeepSeek experience to mean that if it's wrong about a continued AI buildout, the downside for investors won't necessarily be catastrophic. Its strategists reckon that a 20% decline in data-center investment would result in a 3% to 4% hit to next year's S&P 500 earnings-per-share consensus, combined with a reduction of 10% to 13% in its price/earnings ratio. It's nice to know that the highly theoretically downside I'm facing could bring approximate losses that are hypothetically moderate -- but don't hold me to that.

Scott Wren, senior global market strategist at Wells Fargo Investment Institute, says he doesn't find current stock valuations off-putting. He likes industrials and utilities for their side-door exposure to data-center spending, and banks for benefiting from a steepening yield curve. He doesn't recommend diluting the U.S. market's AI concentration by going into small-caps or value funds, or loading up on overseas stocks.

Investors who are sitting on too much cash and waiting for a dip should put a third of it in now, Wren says, and make a plan for adding the rest.

Write to Jack Hough at jack.hough@barrons.com. Follow him on X and subscribe to his Barron's Streetwise podcast.

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

September 26, 2025 17:19 ET (21:19 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.