Key takeaways

  • Momentum for AI stocks (especially big names like Nvidia and Microsoft) has slowed but not evaporated.
  • Many hyperscalers are trading at attractive valuations, and analysts say semiconductors also look cheap.
  • Valuations appear stretched for some of the big “picks and shovels” AI trade winners.

For more than two years, the artificial intelligence trade has perhaps been the most dominant factor driving the bull market in stocks. In recent months, however, the euphoria has faded.

Valuations are looking rich for some AI winners, other high flyers have stumbled, and it’s gotten harder to find overlooked plays. At the same time, worries about an AI bubble are making headlines, a backlash is building against electricity-hungry data centers, and concerns about how some companies will monetize (or even simply pay for) their massive AI investments are dampening sentiment.

Does this mean the AI trade—which swept up stocks ranging from semiconductors to cooling system makers—is played out? Morningstar analysts across sectors touched by the boom say there is more room to run, but investors need to be more discerning.

Such discernment could simply involve waiting for pullbacks in richly valued, volatile names to find more attractive entry points. Or looking deeper into the AI ecosystem at names like Entegris ENTG, which makes specialty cleaning chemicals for the semiconductor industry, and Arista Networks ANET, which sells networking equipment like Ethernet switches to data centers.

“There’s a long runway for spending to keep going,” says Morningstar senior equity analyst William Kerwin, who covers technology firms in the AI pipeline. “If there’s a compelling narrative around the revenue generation, I think it has a ton of legs.”

What ss the AI trade?

At the heart of the AI trade are technology stocks. When AI mania began to take hold in the market in 2023, shares of chipmakers like Nvidia NVDA were the first to take off. Other semiconductor makers and designers have seen their stocks soar, such as Advanced Micro Devices AMD and Broadcom AVGO. Then there are the hyperscaler firms like Microsoft MSFT, Meta Platforms META, Alphabet GOOGL/GOOG, and Amazon AMZN, which provide the cloud computing resources that AI models require. They are investing heavily in AI infrastructure, especially data centers.

The AI boom soon expanded to what Wall Street calls “picks and shovels” companies—those that supply the tools and infrastructure necessary to make chips and build, power, and maintain AI infrastructure, including the data centers that house the physical servers. Those firms encompass industries as disparate as utilities, construction, and specialty chemical makers. Heating and cooling system companies and natural gas stocks were caught up in the run as well.

After a multi-year rally, investor enthusiasm for AI stocks has cooled, partly thanks to concerns about high valuations and excessive borrowing. The Morningstar Global AI & Big Data Consensus Index is up 28% over the past 12 months but just 1.2% since the start of the fourth quarter of 2025.

Many ‘picks and shovels’ stocks look expensive

Shares of some “picks and shovels” companies soared alongside the chipmakers, and in some cases they now look even more overvalued than their “pure play” counterparts.

Kerwin says that firms with real opportunities to profit from AI don’t always make for good investment opportunities. That disconnect has to do with valuations. He says a real business opportunity might also look significantly overpriced. SanDisk SNDK, which makes memory components for semiconductor chips and has seen a major surge in demand, is trading at a whopping 206% premium.

Another example is Corning GLW, which makes glass, ceramics, and optical fiber for a variety of end markets, including data centers. Investors seized on the name as the AI trade took off, and now its shares are trading at a roughly 54% premium to Morningstar’s fair value estimate. “The stock has gotten far overblown relative to that business opportunity,” Kerwin said late last year. On the other hand, Nvidia (where our analysts also see a compelling and durable business opportunity in AI) is rated 4 stars and is trading at a 22% discount to our fair value estimate.

Demand for AI infrastructure still looks strong

While enthusiasm for the AI trade seems to have cooled, our analysts don’t see demand for semiconductors and other AI bets going away anytime soon, even if growth and stock market returns slow in the years ahead.

Morningstar analysts led by Brian Colello foresee “outstanding demand for AI in the near and medium term,” and they say in their fourth-quarter industry pulse report that recent concerns of a bubble are “unfounded.” They expect that demand to drive strong earnings among semiconductor stocks in the months ahead.

If anything, a silver lining for several months of selloffs in the tech sector has been a cooling of valuations. Many hyperscaler and semiconductor firms are currently rated 4 stars by Morningstar, meaning their stock prices are trading at a discount to our estimates of their intrinsic values. Our analysts believe the median semiconductor stock was 6% undervalued as of mid-January. Broadcom is rated 4 stars and trading at a 32% discount. On the hyperscaler side, Microsoft is rated 4 stars and is trading at a 25% discount. Meta is similarly undervalued and is rated 4 stars.

Beyond the major semiconductor and hyperscaler firms, data infrastructure and software companies are some of the most visible components of the AI trade. These firms help manage and store all the extra data that AI programs need, as well as make the physical components for those storage systems and chips. Kerwin says his top picks in this category are chip designer Marvell Technology MRVL, semiconductor giant Broadcom, and Arista. Over the long term, he says these companies will likely benefit from strong demand tailwinds.

“Infrastructure demand is extremely real and I think very durable, because …all the ‘picks and shovels’ are out of stock. They are fully sold out for at least the next year, if not the next two years already,” Kerwin said late last year. “We have confidence that everything will stay supply-constrained through 2026, which should help drive continuing good results and stock performance,” he added recently.

Industrial heating and cooling, chemicals

Beyond semis, much of the AI trade is related to data centers. Nicholas Lieb, who covers industrials, likens them to “massive digital brains.” Contained within this category are specialized companies that handle heat and cooling and power management equipment. Lieb explains that such products have a much slower pace of development, as well as longer lifecycles.

He says this means firms like Vertiv VRT, Eaton ETN, and Wesco International WCC can be a safer way for investors to get exposure to AI and data centers. Their business cycle tends to be less volatile and carry a lower risk of disruption. Of those three stocks, Lieb says Vertiv’s revenue is most exposed to the AI buildout.

While there is always a risk that spending on AI could slow as data centers reach capacity, Lieb says that isn’t the case now. “Spending is ramped up,” he says, noting that Eaton’s data center distribution business is growing 65% per year.

But while these firms might be safer for investors, they’re not necessarily cheaper. “It seems like the market has already recognized their exposure to AI,” Lieb says. Vertiv and Wesco are trading in 2-star territory, while Eaton is rated 3 stars. He notes, however, that stocks that aren’t necessarily “cheap” can still generate good returns, as long as their earnings compound at a steady clip. Take Vertiv. Lieb expects earnings per share growth to compound at a rate of 20% per year, contributing to a projected annual return of 15% for investors, even accounting for the drag from the stock’s elevated valuation.

In a similar category are specialty chemical companies. According toSeth Goldstein, who covers the basic materials sector for Morningstar,these are also poised to benefit from robust demand for data centers. He calls out Qnity Electronics Q, which has 90% of its revenue tied up in semiconductor and electronics production. Shares shot up after the firm was spun off from DuPont and quickly posted strong results related to the semiconductor business. Our analysts consider Qnity fairly valued.

Goldstein also highlights chemical firm Entegris ENTG, which counts Taiwan Semiconductor Manufacturing TSM as its largest customer. Shares are trading in overvalued territory, but were priced at a sizable discount just a month or two ago, when AI names sold off across the board. Because the firm’s fortunes are so intertwined with AI and semiconductor growth, “market sentiment plays a big role and keeps the stock pretty volatile,” Goldstein explains.

Power grab

Utilities stocks surged nearly 20% in 2025, and Morningstar analysts say that momentum is likely to continue this year as more AI data centers (which require massive amounts of electricity) are built. In a recent report, Morningstar analysts led by Travis Miller say that buildout represents a “transformational” source of demand growth for the industry.

But it’s a mixed story on valuations. Overall, utilities are on the high side, albeit below peak levels thanks to a fourth-quarter selloff. The utilities sector was trading at a premium of nearly 20% last fall, but as of late January, that premium had dropped to 5%.

One of Miller’s team’s top picks is Alliant Energy LNT, which counts four data centers responsible for 12% annual sales growth among its customers. Alliant is rated 4 stars and is trading at a 6% discount to its fair value estimate.

Natural gas companies like Williams WMB and Energy Transfer ET are also poised to benefit from the data center buildout, according to Morningstar analyst Adam Baker. Williams operates a key pipeline on the East Coast and has key gathering and processing operations in Appalachia, while Energy Transfer has gathering and processing operations as well as transmission lines in Texas that will allow it serve data centers there.

AI in practice

In addition to companies directly within the AI supply chain, there are those that could benefit from its use. Goldstein points to Corteva CTVA, an agricultural firm spun out of DowDuPont that is using AI to speed up its development of new seeds. The stock is trading at an 11% discount and is rated 4 stars.

Karen Andersen, who covers healthcare stocks for Morningstar, sees a similar phenomenon unfolding as firms leverage or build AI tools to help with drug development. Eli Lilly LLY, for example, recently announced a new initiative with Nvidia that will use AI to accelerate drug discovery. Roche RHHVF and Amgen AMGN have announced similar plans.

However, Anderson notes that it will be years before any productivity gains materialize for these firms. “It’s hard to see what the ultimate outcome is going to be,” she says. “It’s not clear to me that this has been factored into share prices.”

Be discerning and watch for dislocations

With markets so closely attuned to AI winners and losers, can investors still find good buys? Analysts say the key lies in identifying gaps between share prices and valuations, even if valuations never look dirt-cheap.

Kerwin points to Arista, which delivered strong earnings last quarter, prompting our analysts to raise their fair value estimate of the stock. But amid a broader selloff in tech and AI names, Arist shares sold off sharply. Kerwin says dislocations like this, in which prices fall based on uncertainty rather than fundamentals, can constitute good entry points. “If you can get a ‘fair’ price, even if its not a steal, that constitutes a pretty good deal in this market,” says Kerwin.

Get Morningstar’s insights in your inbox