Regulated Utilities vs. Merchant Power: A Sector Split

Picture a June afternoon when a heat dome parks itself over Texas. Air conditioners across ERCOT pull the grid toward its limit, wholesale power prices that normally sit in the $30-per-megawatt-hour range spike past $2,000, and a gas-fired plant that was barely profitable in April prints more margin in six hours than it did across the entire previous quarter. A few states over, a regulated utility serving a nearly identical heat wave earns roughly the same return it was always going to earn. Its allowed return on equity doesn't spike with the thermometer. It was set months or years earlier, in a rate case, by people who have never operated a power plant.

That contrast is the whole story of the power sector, and most retail investors miss it because "utility stock" gets used as if it described one business. It doesn't. A regulated utility and a merchant power producer both generate and sell electricity, but they operate under almost opposite economic logics — one earns a contractually bounded return regardless of how hot the summer gets or how tight the market is; the other earns whatever the market clears at, for better and for worse. Confusing the two is how an investor ends up sizing a defensive-utility position in a business that behaves like a commodity cyclical, or the other way around.

I'll separate the two models cleanly here, walk through where a handful of large-cap names actually sit on that spectrum — several of them straddle it, which is its own lesson — and get into what's changing now that data centers are rewriting electricity demand forecasts for the first time in two decades.

The Regulatory Compact: How Regulated Utilities Actually Earn Their Return

Start with the regulated side, because its economics are genuinely unusual relative to the rest of the market. A regulated utility doesn't earn a return by selling electricity at a profitable spread over its cost of producing it, the way almost every other business earns money. It earns a return by being granted one, through a rate case, on the capital it has invested to serve its territory — its "rate base." A public utility commission sets an allowed return on equity, the utility charges rates designed to hit that return, and the whole relationship gets periodically renegotiated in a proceeding that looks more like a legal settlement than a competitive sale.

Georgia Power's most recent alternate rate plan, reached in a settlement approved by the Georgia Public Service Commission in December 2022 and covering 2023 through 2025, set the company's authorized return on equity at 10.5% on a 56% equity ratio — with an earnings-sharing mechanism above that band that returns some of any excess to customers rather than letting it flow through entirely to shareholders. Florida Power & Light's situation shows the same structure moving with market conditions: its four-year settlement, approved by Florida regulators in October 2021, was later adjusted by the state commission in September 2022 to raise the authorized ROE range to 9.8%–11.8%, a response to rising Treasury yields shifting what regulators consider a fair return on utility capital — a concept closely related to a company's cost of capital, except set by a commission rather than discovered in a market. Neither number moved because Georgia or Florida had an unusually hot or cold year. They moved because a legal and political process, not a market, set the price.

Merchant Power: Selling Into a Market That Owes You Nothing

Merchant generators operate under the opposite logic entirely. They build or acquire power plants, sell that output into a wholesale market — PJM, ERCOT, or one of the other regional grid operators — and collect whatever the market clears at that hour, that day, that year. No rate case. No allowed ROE. No earnings-sharing mechanism smoothing the outcome. And no floor, either, when demand slackens or a plant sits uneconomic for a season — which moves merchant earnings closer to a cyclical business than the sector's defensive reputation implies.

Vistra Corp is the clearest large-cap expression of this model. Prior to closing its acquisition of Energy Harbor on March 1, 2024, Vistra operated roughly 37,000 megawatts of generation capacity almost entirely outside rate regulation — natural gas, coal, nuclear, solar, and battery storage sold competitively into whichever market each asset sits in. The Energy Harbor deal added about 4,000 megawatts of nuclear capacity and combined it with Vistra's existing nuclear and renewable assets to form Vistra Vision, a roughly 7,800-megawatt zero-carbon generation and retail platform serving close to 5 million retail customers. Constellation Energy, spun off from Exelon in 2022, runs the largest competitive nuclear fleet in the country — about 22 gigawatts across 14 stations and 25 reactor units, which generated 174 terawatt-hours of zero-emissions electricity in 2023 at a fleet-wide capacity factor of 94.4%. That's an extraordinary operating statistic. It has almost nothing to do with a regulatory settlement, though — Constellation earns on that fleet by selling power and, increasingly, by signing long-term contracts directly with buyers, not by petitioning a commission.

Three Business Models Wearing One Sector Label

Group the large-cap names in this space by what they actually earn their returns on, and a clean industry structure emerges — one that a market-cap-weighted utility index or a simple dividend-yield screen obscures entirely.

• Pure regulated incumbents: Southern Company (through subsidiaries Georgia Power and Alabama Power) and American Electric Power. These companies earn essentially all of their returns through allowed-ROE rate base growth — transmission buildout, generation upgrades, grid hardening — approved state by state. Growth here is a function of capital investment approved by regulators, not of selling more electrons at a better margin.
• Hybrid utilities with a competitive arm: NextEra Energy is the defining example. Florida Power & Light is a purely regulated utility subject to the ROE mechanics above, while NextEra Energy Resources, its competitive arm, develops and operates wind, solar, and battery storage projects sold under long-term contracts or directly into wholesale markets. The two segments answer to entirely different economic logics inside one holding company, which is exactly why NextEra doesn't sort cleanly into either bucket.
• Pure competitive generators: Vistra Corp and Constellation Energy. Both are almost entirely exposed to wholesale power prices and market design, with no allowed-ROE mechanism smoothing the outcome. Their earnings move with commodity prices, weather, and — increasingly — the terms of the direct power-purchase contracts they can negotiate with large individual buyers.

NRG Energy sits in a fourth, messier category worth naming separately: a merchant generator that also owns a large retail electricity supply business, which changes its risk profile again. Retail margins can partially offset generation-side commodity swings, but the company still carries none of the rate-base-and-allowed-ROE structure anchoring the regulated names.

What's Actually Changing: Data Centers and the First Real Demand Growth in Twenty Years

US electricity demand was essentially flat for two decades, from the mid-1990s efficiency gains through the mid-2010s, which is part of why utility investing became synonymous with slow, bond-like compounding. That's the backdrop AI data center demand is disrupting — part of the broader energy transition reshaping how power gets generated and priced — and it's disrupting the regulated and merchant sides differently.

For regulated utilities, new large-load demand — a data center campus requesting hundreds of megawatts of interconnection — becomes an opportunity to expand rate base: new transmission lines, new generation, new substations, all approved through the same rate-case mechanism and all earning the same allowed ROE on a larger capital base. That's a secular tailwind moving through a familiar regulatory channel. For merchant generators, the same demand shows up as a buyer willing to sign a long-term power-purchase agreement directly with a specific plant — the kind of arrangement that let a shuttered nuclear unit at Constellation's Three Mile Island site restart under a 20-year agreement with Microsoft, announced in September 2024. That's not rate-base growth. It's a merchant asset locking in a contracted, bond-like cash flow that starts to resemble the regulated model's predictability, minus the regulator.

I'm less confident than I'd like to be about how much of this data-center demand growth will actually show up as forecast, and by when. Interconnection queues are long, utility load forecasts have been wrong in both directions before, and some announced data center campuses get delayed, downsized, or cancelled entirely once financing or chip supply gets complicated. But the direction of the shift — toward a sector that hasn't had genuine demand growth to argue about in a generation, suddenly having some — looks real enough that it's already reshaping capital allocation decisions across both models.

That uncertainty cuts both ways on the market-design side, too. PJM's board proposed in late July 2026 to curtail power to large new loads — including data centers — during grid emergencies starting June 1, 2027, specifically targeting facilities that haven't secured their own generation or contracted supply, after the grid operator's most recent capacity auction came in roughly 6.8 gigawatts short for the delivery year beginning mid-2028. That's the concrete mechanism behind why a data center operator increasingly would rather sign a long-term contract with a specific power plant than assume the regional grid will simply have capacity waiting for it.

Risks Each Model Carries That the Other Doesn't

Regulated utilities carry regulatory and political risk almost exclusively: a hostile commission appointment cycle, a state legislature capping rate increases, or a disallowed capital project — built, but ruled imprudent, and therefore excluded from rate base — can impair returns in ways no amount of operational excellence offsets. Wildfire liability has become the sharpest version of this risk in the West. A regulated utility found to have caused a catastrophic fire can face liabilities that dwarf a normal year's allowed return, regardless of how well-run the rest of the business is.

Merchant generators carry commodity and market-design risk instead. Natural gas prices, capacity market rule changes — PJM's capacity auction reforms have moved cleared prices sharply in both directions over the past few years — and plant-level operational risk all flow directly to earnings with no rate case standing between a bad year and the bottom line. And leverage matters more here: a merchant generator financed like a regulated utility, with a similar debt load but none of the guaranteed-return cushion, carries a meaningfully riskier balance sheet than the sector's defensive reputation suggests to an investor who hasn't separated the two models. The scale advantages that protect a regulated incumbent's return don't translate the same way into a merchant generator's earnings, either — scale there just means more megawatts exposed to the same market swings.

Key Takeaways

• Regulated utilities earn a contractually set return on rate base, approved through state rate cases. Georgia Power's 10.5% authorized ROE (2023–2025 settlement) and FPL's 9.8%–11.8% range (adjusted September 2022) show that number is a regulatory and political outcome, not a market one.
• Merchant generators like Vistra Corp (roughly 37,000 MW pre-Energy Harbor, expanding into the roughly 7,800-MW Vistra Vision platform) and Constellation Energy (about 22 GW of nuclear capacity, a 94.4% capacity factor in 2023) earn whatever the wholesale market clears — no floor, no allowed return.
• Hybrid companies like NextEra Energy run both models under one roof, which is why treating "utilities" as a single defensive sector obscures more than it reveals.
• Data center demand is a secular tailwind reaching both models through different channels — rate-base growth for regulated names, direct long-term power-purchase contracts for merchant ones — but the pace and durability of that demand remains one of the more genuinely uncertain forecasts in the sector right now.