
Data center electricity demand has outpaced every grid forecast. Constellation Energy's nuclear fleet and GE Vernova's power hardware backlog are the two clean ways to play the gap.
Data center electricity demand has grown faster than any forecast the grid operators published three years ago. That has created a two-sided opportunity for the power sector: companies that generate the electricity and those that build the equipment to distribute it.
Hyperscalers have committed hundreds of billions of dollars to new computing clusters. Those clusters need power 24 hours a day, seven days a week, and preferably from sources that do not add carbon to the grid. That combination has made nuclear the preferred solution for the largest build-outs.
Constellation Energy is the biggest U.S. nuclear operator by capacity, a position that has turned its fleet into a bottleneck asset for AI expansion. In 2024 the company signed a 20-year power purchase agreement with Microsoft to restart the Three Mile Island Unit 1 reactor in Pennsylvania, now renamed the Crane Clean Energy Center. The deal is the longest fixed-price nuclear PPA on record and covers the plant's entire output.
Microsoft's willingness to lock in a two-decade contract shows how tightly data center planning is tied to power availability. Hyperscalers are spending tens of billions on new campuses, but each site is worthless if the local utility cannot deliver enough electrons. Constellation's existing reactors, running at over 93% capacity factors, offer something wind and solar cannot: guaranteed round-the-clock output without storage.
The economics have shifted in Constellation's favor. Regulated utilities have been slow to build new generation, and interconnection queues for renewables stretch past 2030. Constellation's existing nuclear fleet can serve load immediately, and the company is pursuing license renewals that would extend plant lives to 80 years. The stock has fallen more than 25% year to date through Aug. 20, which has put the forward earnings multiple below its five-year average. That has made the shares cheaper than they have been since the AI trade first lifted utility stocks in early 2024.
GE Vernova occupies a different part of the same supply chain. It does not generate power. It sells the hardware that makes power usable: gas turbines, high-voltage transformers, switchgear, and grid control systems. Data center developers are buying all of it.
The company said data center power equipment generated $5 billion in orders during the first half of 2026, more than double the full-year total for 2025. The broader backlog has swelled to $176 billion, a level that effectively gives the company a multiyear revenue floor. Most of those orders come with cancellation penalties, so the backlog is not soft.
GE Vernova's gas turbines are its strongest product line. Natural gas is not carbon-free, but it is dispatchable and cheap enough that hyperscalers are using it as a bridge fuel while new nuclear projects work through licensing. The company also makes grid interconnection equipment that every data center needs to sync with the local utility. That part of the business tends to be overlooked but accounts for a growing share of revenue.
The stock has more than doubled since the start of 2025, which has raised questions about how much is already priced in. But the company's position in the supply chain is structural. Every new data center needs a transformer and a switchyard. There are only a handful of suppliers that can produce them at scale, and GE Vernova is the only one that also builds the gas turbines that run when solar is not available.
Neither stock is cheap on a historical basis. Constellation trades at roughly 22 times forward earnings; GE Vernova trades near 35 times. But the demand driver is not cyclical. It is tied to a capital spending cycle that hyperscalers have already locked in through 2030. The question is not whether orders will come. It is whether the supply side can keep up.
Prepared with AlphaScala editorial tooling from the source reporting linked above. Indexable analysis may include a cited Alpha Score value. Publishing checks screen each story before release. Educational coverage, not personalized advice.