The hottest AI stock of 2026 might not be a chipmaker at all, according to a recent quiet joke among New York portfolio managers. It could be a business that manufactures gas turbines in upstate New York or one that produces switchgear in a plant in Wisconsin where the parking lot is still full at six in the morning. Before you see the tape, it sounds ridiculous. While the major software names take turns collapsing, companies like GE Vernova, Caterpillar, and Eaton—names your grandfather may have owned—are subtly circling the wider market.
Calling this a rotation, classifying it as a “value comeback,” and moving on would be tempting. However, it took me some time to realize that something more specific was happening. One thing unites the industrial stocks that are currently outperforming the market, and it has nothing to do with margins, management, or even valuation. It’s that each of them has, in some strange, unglamorous way, turned into a component of the tangible framework supporting the AI boom. The turbines. The transformers. the systems for cooling. The trucks are transporting steel to a partially constructed data center near a Virginia interstate exit. The trade is that.
It’s possible that the market took longer than necessary to come to this conclusion. Software, models, and the chips that powered them dominated the AI discourse for the majority of 2024 and early 2025. After that, the electricity bills became due. There are hundreds of hyperscale data centers planned, each capable of drawing as much power as a small city. The grid that keeps them operating needs to be built by someone, and there aren’t as many businesses that can do it as people think.
GE Vernova is now clearly the poster child. Three years ago, no one would have predicted that its order book for natural gas turbines would extend into the late 2020s. The boring, necessary, and backordered electrical parts that are located between the substation and the server rack are sold by Eaton. Dirt for the foundations is being moved by Caterpillar’s machinery. Because so much of what they distribute ends up bolted into a factory floor somewhere, even smaller names like Applied Industrial Technologies are catching the updraft.
Duration is what sets this apart from a normal capex cycle. Spending doesn’t fluctuate on a quarterly basis. Data centers, grid resilience, and the costly and time-consuming task of rebuilding an American manufacturing base are all part of this multi-year commitment. It includes incentives from the government. So is the straightforward fact that an AI model requires copper, steel, and a functional circuit in order to function. The analysts I respect the most believe that this is more akin to the telecom buildout of the early 2000s than to a typical industrial bounce, but this time the demand is genuine and the clients are solvent.
It’s difficult to ignore the irony as you watch this play out. For twenty years, the dominant narrative maintained that the American economy had transcended all of this, including factories, heavy machinery, and things that spark and hum. The future was meant to be frictionless, digital, and light. Then came the future, which proved to require a lot of concrete.
This is not a guarantee. Cycles shift. Backlogs relax. It’s a strange statement in and of itself, but some of these stocks now trade at multiples that would have appeared rich for a software company a few years ago. Not every name in the index will travel, Graco has stalled, and Resideo has faltered. However, investors don’t seem to want to give up on the larger narrative that the businesses satisfying artificial intelligence’s physical appetite are currently the true AI trade. The question worth considering is whether they ought to.
