The utilities sector has been the quietly best performing sector on the S and P 500 so far in 2026. Through Friday April 17 the Utilities Select Sector SPDR Fund, ticker XLU, is up approximately fourteen percent year to date, beating the broad S and P 500 by a comfortable margin and beating the Technology sector by a surprising one. The story behind that outperformance is not the traditional defensive rotation narrative. The story is AI power demand, and it took investors longer than it should have to start pricing it correctly.
The setup matters. For roughly a decade preceding 2023 the utilities sector traded as a pure bond proxy. Investors bought utilities for yield, rotated in and out based on the ten year Treasury, and generally treated the sector as a slow growth dividend play. Earnings growth across the sector was low single digits. Capital expenditure was focused on grid maintenance and slow renewable transition. There was no growth story.
That pattern broke starting in early 2024 as the first serious estimates of AI data center power demand began to hit investor models. The estimates were shocking. Projections for US data center electricity consumption doubled in roughly eighteen months, then doubled again. PJM Interconnection, the grid operator serving thirteen states and Washington DC, now projects data center load growth at over forty percent through 2030. ERCOT, the Texas grid, is projecting more than sixty two gigawatts of new data center demand over the same window. The math looked unsustainable on paper, but the capital was already moving.
The companies that captured the first phase of the rally were the independent power producers and the utilities with direct exposure to data center heavy service territories. Vistra, Constellation Energy, and NRG all ran hard through 2024 and 2025. Constellation in particular posted returns in excess of one hundred fifty percent over the eighteen month window leading into the end of 2025, as investors bid up its nuclear fleet on the thesis that hyperscalers were going to pay premium prices for twenty four seven clean power.
What changed in early 2026 was that the rally broadened. The regulated utilities with no independent power production exposure started outperforming as well. Dominion Energy, which serves the Northern Virginia data center cluster, has now posted a twenty two percent year to date return. Duke Energy is up fifteen percent. Southern Company is up thirteen percent. American Electric Power is up seventeen percent. These are utilities whose regulated returns are set by state public utility commissions, which traditionally capped their upside. The rally suggests investors are now pricing in either faster rate base growth, more favorable regulatory outcomes, or both.
The nuclear sub segment is the most interesting piece. Constellation, Vistra's Comanche Peak assets, Talen Energy, and a handful of smaller nuclear operators have become strategic assets in the AI power story because nuclear provides twenty four seven baseload clean power that hyperscalers increasingly require for their carbon commitments. The direct purchase power agreements that Amazon signed with Talen in 2024 and that Meta and Microsoft have signed with other nuclear operators over the last year are effectively taking nuclear capacity off the public grid and directing it to specific data center campuses. That dynamic creates scarcity pricing in the residual wholesale market.
Small modular reactors are the speculative angle. Several utilities and independent power producers have announced SMR partnerships, and investor interest has spiked on any company with a credible SMR pathway. The problem is that commercial SMR deployment in the US is still several years from first grid connection at scale, which means the stock price moves are running well ahead of the actual revenue. A handful of SMR linked names have been volatile as a result.
The valuation question is the main risk. The utilities sector is now trading at a price to earnings ratio of approximately twenty one, compared to a ten year average of roughly seventeen. The yield spread versus the ten year Treasury has compressed to historical narrows. If investor expectations for data center load growth soften, or if regulatory outcomes on rate cases come in worse than expected, the sector has room to give back meaningful gains. That said, the pace of new data center announcements is still accelerating, which continues to validate the underlying thesis.
Dividend investors still have a role in the sector but the play has shifted. The old utilities thesis of slow growth plus steady yield is being replaced by a faster growth plus moderate yield thesis. Payout ratios have declined modestly as earnings growth accelerates. Capex plans have roughly doubled across the major regulated utilities over the last two years, which means more invested capital earning regulated returns in future years.
The practical investor takeaway is that the sector deserves a different analytical frame than it had two years ago. The right comparison set for a utility like Dominion in 2026 is less a bond proxy and more a regulated infrastructure growth business. That reframing is how investors make sense of the current valuations without concluding the whole sector is in a bubble.
Watch the Q1 earnings reports this month. Guidance revisions are the signal to track.