#Is Buying a Utility a Safer Way to Invest in AI?
AI runs on electricity, so utility stocks have become one of the more popular ways to invest in the AI buildout without buying chipmakers or tech giants directly. Many investors see them as the steadier option. That is partly true, but it depends heavily on which kind of utility.
The sector splits into two camps that react to AI news in very different ways. The gap shows up most clearly when sentiment turns. In November 2026, Democratic wins in the US midterm elections raised fears of tighter limits on data center construction, and AI stocks sold off sharply [1]. Power producers most exposed to data center demand, such as Vistra (NYSE: VST) and Constellation Energy (NASDAQ: CEG), were among the companies caught in the selling [2].
How hard a utility stock gets hit comes down to how the company gets paid. But first it helps to see how much of the power story rests on AI spending.
#AI Spending Is Driving the US Power Buildout
The US is in the middle of a capital spending cycle dominated by AI. Large tech companies are pouring money into chips, servers and the data centers that house them. ING estimates that AI and data center investment accounts for about a third of US economic growth in 2026, and that data centers and power plants are the only parts of US construction where spending is growing [3].
Utilities sit downstream of that spending. Every new data center needs a reliable power supply, so tech budgets turn into requests for new power plants, substations and transmission lines. That makes the utility opportunity real, but it also ties it to decisions made in Big Tech boardrooms. If the largest tech companies slow their spending, fewer data centers get built and the demand forecasts behind utility investment plans shrink with them.
A sharper way to see the markets in just 5 minutes.
Same news, different lens. We cut through the noise and hand you the overlooked ideas and the deeper read the crowd misses. Join 38,000+ investors seeing the markets differently.
#How Regulated Utilities Make Money From Data Centers
A regulated utility is a local monopoly. In return for that monopoly, a state commission sets the prices it can charge. The utility earns an agreed profit, called an allowed return, on the grid and power plants it builds. The total value of those assets is known as the rate base.
That means a regulated utility makes more money mainly by investing more, not by selling more power at a higher price. Data centers matter because they justify that investment. A large new data center needs new power plants, substations and transmission lines, and each one adds to the rate base.
Southern Company (NYSE: SO), Duke Energy (NYSE: DUK) and DTE Energy (NYSE: DTE) are examples of this model. DTE, which serves southeast Michigan, has signed agreements to supply about 1 gigawatt to a Google data center and 1.4 gigawatts to Oracle, and has expanded its five-year investment plan to match [4].
#Why Power Producers Swing Harder With AI Sentiment
Independent power producers work differently. They own power plants and sell the electricity into wholesale markets or through contracts with big buyers, including tech companies.
When demand is strong and power prices rise, their profits can grow much faster than a regulated utility's. That is why they became some of the most popular AI trades. The flip side is that their share prices respond quickly to any hint that data center growth could slow. Investors are pricing in future demand that has not arrived yet, so a change in expectations moves the stock.
The line between the two camps is not always sharp. Some power producers now sign long-term deals directly with tech companies, which locks in part of their income. Constellation, for example, agreed a 20-year deal to supply Microsoft by restarting a reactor at Pennsylvania's Three Mile Island plant [5]. Contracts like that make a producer's earnings more predictable, though the share price can still move with wider AI sentiment.
The scale of expected demand is large. Data centers used about 4.7% of US electricity in 2024. Lawrence Berkeley National Laboratory projects that share could reach about 11.8% by 2030, within a wide range of 9.5% to 15.3% [6]. The width of that range is itself the point. Nobody knows how much of it will be built.
#Regulated Utilities vs Power Producers at a Glance
Regulated utilities | Independent power producers | |
|---|---|---|
How they earn money | An approved return on the grid and plants they build | Selling power at market prices or under contract |
Who sets prices | State regulators | Wholesale markets and contract terms |
Upside from AI demand | Steady, tied to how much they invest | Higher, tied to power prices and demand |
Reaction when AI sentiment turns | Usually milder | Often sharp |
Main risks | Tougher regulators, funding needs, bill backlash | Weaker demand, falling power prices |
Examples | Southern Company, Duke Energy, DTE Energy | Vistra, Constellation Energy |
#A Third Way In Through Grid Equipment
Utilities are not the only route into AI power demand. A third group supplies the picks and shovels. These are companies that build transmission lines or make the cooling and electrical gear inside data centers, such as Quanta Services (NYSE: PWR) and Vertiv (NYSE: VRT).
They earn money while the building happens, whoever ends up selling the power. That also makes them among the most sensitive to any slowdown in data center construction, so they tend to sit closer to the power producers than to regulated utilities in how they trade.
#Are Utility Stocks Defensive When AI Stocks Fall?
Regulated utilities have a reputation as defensive stocks, meaning shares that tend to hold up when the wider market falls. Most of their demand comes from millions of homes and businesses that keep using power whatever happens to tech stocks, and data centers are one growing part of that.
That reputation has limits. Utilities often pay steady dividends, so investors compare them with bonds. When interest rates rise, new bonds offer a higher yield, meaning a bigger return for holding them, so utility shares can look less attractive by comparison. A utility can be shielded from an AI sell-off and still fall when interest rates jump.
Dividends deserve a closer look too, since they are often the reason investors own utilities in the first place. A company spending at record levels needs cash for new plants and grid upgrades. That can slow dividend growth, or push the company to borrow more and issue new shares to keep paying out.
#Where the Regulated Model Can Break Down
The regulated model only works if regulators keep approving the spending and the prices. That is a political decision, and power bills have become a political issue. The average US household paid about 16.5 cents per kilowatt-hour for electricity in 2024, and the government's energy statistics agency expects that to reach about 18.2 cents in 2026, roughly a tenth higher in two years [7].
In the November 2026 election, Democrats won control of Georgia's Public Service Commission, which regulates Southern's Georgia Power unit, for the first time in two decades. Candidates had campaigned against a run of rate increases. Michigan voters also approved a ban on regulated utilities donating directly to state politicians [8]. Tougher regulators can approve smaller rate increases or cut the allowed return, which slows profit growth.
There are other risks too. The biggest sits upstream. If tech companies cut their AI budgets, data center demand could come in below forecasts, and a utility may have built for customers who never arrive. Large investment plans are often funded by borrowing and issuing new shares, which spreads future profits across more stock. And if household bills rise to pay for data center growth, the political backlash tends to land on the utility. Supply can be a constraint as well, as the power squeeze in the western US shows, where drought and retiring plants are straining the grid just as demand climbs.
#Five Questions to Ask About a Utility's Data Center Exposure
For investors researching a utility as an AI-linked holding, these questions help separate the steady stories from the stretched ones.
How much of the company's expected growth depends on data centers, and how many of those depend on a handful of tech companies?
Do data center customers sign contracts that make them pay even if they use less power than planned, protecting other customers from the cost?
How supportive are the company's regulators, and when is its next rate case, the formal process for setting prices?
How will the investment plan be paid for, and could it slow dividend growth or lead to many new shares being issued?
How much of the business sits outside regulation, where profits move with market prices?
The answers usually show up in a company's investor presentations and annual reports, and in local news coverage of its rate cases. The signals worth watching across the sector are Big Tech's spending plans, which the largest companies update with their quarterly results, new limits on data center construction, the first decisions of newly elected regulators, and how fast household power bills keep rising.