Vertiv and GE Vernova suggest AI's money is moving to power
Vertiv's 61% sales growth and GE Vernova's near-$200 billion backlog suggest AI spending is flowing to power and cooling suppliers, but position size matters more than the story.
Photo: David E. Lucas / Wikimedia Commons (Public domain)
What if the best-run AI business isn't building chips or models, but keeping the lights on and the servers cool? Wednesday's earnings made that case better than any keynote could.
Vertiv reported third-quarter net sales growth of 61%, organic orders growth of over 100% and diluted EPS growth of 91% (78% on an adjusted basis). It also raised its full-year 2026 guidance across all key metrics. The same day, GE Vernova posted EPS of $4.81 against a $4.03 consensus, on revenue of $12.64 billion. Its CEO said the company now expects its backlog to exceed $200 billion in the fourth quarter, with at least 130 GW of gas equipment under contract by the end of 2026.
The bottleneck is now physical
For two years the AI conversation was about who had the best silicon. These numbers suggest the scarce resource may have shifted. A data center needs electricity first and a way to get rid of the heat second. Orders more than doubling at Vertiv, and a backlog at GE Vernova that the call transcript puts at $197 billion, are what demand for those things looks like when it shows up in contracts rather than in slide decks. To be clear, the companies talk about demand for electricity infrastructure; the AI link is our reading, not something their results spell out.
We think this is the underrated part of the story. Chip buyers may be able to change their minds or their suppliers relatively quickly, while turbines, switchgear and cooling systems tend to be planned and ordered well ahead. A backlog is business a company has already contracted but not yet delivered, and in our view that is a sturdier signal than one hot quarter. It also changes who the winners are: the companies that can actually build and ship heavy equipment, not only the ones with the best narrative.
The market's mood is visible elsewhere too. Littelfuse, an electronic component provider, saw revenue rise 22.7% to $766.7 million. It missed analyst expectations on revenue, yet the stock still rose 7.7% after its guidance came in above consensus. Its CEO credited broad-based demand strength, not AI specifically, so it's a looser fit for the theme. Still, when even a revenue miss gets rewarded, it looks like investors are paying for strong industrial demand as much as for the quarter itself.
Great numbers are probably already in the price
Here is the uncomfortable part. A 61% sales gain and a raised outlook are the kind of results investors hope for, and when the good news arrives all at once there is little left to surprise anyone. These are not sleepy utilities. In our view they trade on expectations, and expectations leave little room for even a modest disappointment.
There's also a quieter risk, which is that everyone now knows the story. GE Vernova has repurchased 15.6 million shares for $9.9 billion since December 2024, at an average price of approximately $635 per share, according to its own release. That says the company believes in its shares, but arguably it also shows how far the price has traveled. A great business and a great entry price are not the same thing, and strong earnings don't turn the second into the first.
We'd also be wary of treating "AI power" as one trade. Vertiv, GE Vernova and Littelfuse sell different products to different customers, and Littelfuse still draws 23.8% of its revenue from automotive. Each of these names can surprise in its own direction.
What we'd do with this, and what we wouldn't
This is not personal advice, and nothing here is a forecast. But the logic of the thesis points to a few habits. Start by treating these stocks as a satellite holding rather than the core of a portfolio. Then size each position so that a sharp one-day drop would be annoying, not damaging. And if you're buying with expectations this high, consider building in stages rather than all at once, so one earnings reaction doesn't set your whole cost basis.
The verdict: the shift toward power and cooling looks real, it is backed by orders, and it deserves attention. But the evidence that the business is excellent is also the reason the margin for error is thin. In a trade this crowded, the discipline is in the position size, not the conviction.
No topics yet: start the first one.
More stories
Deutsche Börse is a toll booth on volatility, not a hedge against it
Volatility feeds exchange operators, and Deutsche Börse's 40% one-year rally shows investors know it. But a price near its yearly high means the good news may already be in the quote.
The new American Dream is a brokerage account, and that's the problem
Stocks have overtaken housing as America's top wealth driver for the first time since WWII. We think that's less a triumph than a concentration risk, unevenly shared.
Record stock exposure isn't a victory lap: it's a risk worth checking
Stocks hit 48.2% of Americans' financial assets, above the dot-com peak. Rallies feel great, but that concentration is a risk to check against your own horizon.
Norway's giant fund lost on bonds: "safe" never meant "stable"
The world's largest sovereign fund posted a $29.7 billion quarterly loss as bonds slid, and the lesson for ordinary savers is about what bonds are actually for.
