Same Wire, Half the Return: Why a Regulator, Not the AI Boom, Decides What a Utility Earns

In the summer of 2026 Duke Energy unveiled the largest capital plan any regulated U.S. utility has ever put on a slide: 103 billion dollars through 2030. Rate base, the company says, will grow seven to nine percent a year. Earnings per share? Five to seven percent. Two full percentage points evaporate somewhere between the transmission line and the shareholder — and that gap is not a rounding error. It is the entire business model in one number.

Almost every utility bought in 2026 is bought on an artificial-intelligence story. Data centers need power, power needs wires, wires belong to utilities. The logic is not wrong, but it skips the one step where it is decided how much of the boom reaches the owner of the stock. Because a regulated utility does not make money where the narrative assumes it does.

The only formula that matters

An ordinary company earns price times volume minus cost. A regulated utility earns something structurally different, and it fits on a single line:

Earnings ≈ rate base × allowed equity ratio × allowed return on equity

The rate base is the book value of assets a commission has deemed used and useful. On that number, and only on that number, the utility is permitted to earn a return. Not on revenue. Not on kilowatt-hours delivered. Not on the market value of the asset. On the regulatory book value.

Three consequences follow, and together they explain the entire sector. Growth comes from spending, not from selling. The level of the return is set by a commission, not negotiated with a customer. And between the moment the money is spent and the moment it starts earning inside an approved tariff, time passes — time that in this business costs real money.

Why the kilowatt-hour barely touches profit

This is where most investors model the sector incorrectly. At a fully regulated utility, fuel and purchased-power costs are typically a pass-through: they move to the customer dollar for dollar through adjustment clauses, with no markup. Gas prices rise, the household bill rises, utility profit does not move. Gas prices fall, the same thing happens in reverse.

In many states the decoupling goes further still. Under revenue decoupling mechanisms, a utility’s allowed revenue is separated from the volume it actually sells: sell more and it refunds the difference, sell less and it recovers it. The original purpose was to stop energy-efficiency programs from working against the utility’s own interest. The side effect for the investor is severe — volume growth as such has become earnings-neutral.

Germany arrives at the same place by a different road. Under incentive regulation, network operators run against a revenue cap that is essentially fixed for the length of a regulatory period and derived from cost audits, not from energy transported. Push more electricity through the same wire and the per-kilowatt-hour charge tends to fall; profit does not rise.

That leaves shareholders exactly one lever: the rate base. And the rate base only grows when capital goes into the ground, up the towers and inside the substations.

What the AI boom actually triggered: a capex wave

Which is why the right metric for the AI boom is not electricity consumption but capital expenditure. And on that measure the numbers are genuinely large. According to the Edison Electric Institute, the association of U.S. investor-owned utilities, member companies invested roughly 208 billion dollars in 2025 — the highest figure ever recorded. For 2026 the plan is about 238.8 billion dollars, a jump of roughly 17 percent. Within a single year the five-year plan was revised upward from 1.1 trillion dollars for 2025 through 2029 to 1.4 trillion dollars for 2026 through 2030.

The association also maintains a running snapshot of large-customer work: roughly 900 billion dollars of data center and large-load investment supported by member companies, representing more than 55 gigawatts of connected load. For scale, that is comparable to the entire installed generating capacity of Germany.

At the company level it looks like this:

Company	Capital plan	Rate base growth p.a.	EPS growth p.a.	Gap
Duke Energy	$103bn through 2030	7–9% through 2029	5–7% through 2030	about 2 points
Southern Company	$81bn 2026–2030	9% (state-regulated)	7–8% (long term)	about 1.5 points
NextEra Energy	about $75bn	—	—	—
Dominion Energy	about $50bn	—	—	—
E.ON	€48bn through 2030	about 9%	dividend up to +5% through 2028	about 4 points

That last column is the real subject of this analysis. Both numbers appear in every investor deck, side by side, on the same slide, and almost nobody asks why they diverge.

The wedge: where the two points go

If rate base compounds at nine percent and the allowed return is constant, earnings should compound at nine percent too. They do not, and there are three nameable reasons.

Equity dilution. Regulators recognize only a defined share of equity in the capital structure — in the United States typically around 50 percent. In November 2025 Northern Illinois Gas and AEE-Illinois were authorized a 9.6 percent return on a 50 percent equity ratio; PG&E has been authorized 50.50 percent, Southern California Edison 50.00 percent. So expanding rate base by 100 dollars requires roughly 50 dollars of fresh equity. A utility paying out 60 to 75 percent of earnings as dividends cannot fund that from retained earnings. It issues stock. Duke expects roughly 10 billion dollars of common equity issuance between 2027 and 2030. Total earnings rise; they are simply spread across more shares.

The cost of debt. The other half is financed with bonds. But the debt cost embedded in rates is usually a historical blended cost of the existing portfolio. A utility refinancing today at coupons well above the 2010s pays more in cash than it recovers in the tariff — until the next rate case resets the number.

Regulatory lag. This is the most underrated line item of the three. Between the day an asset enters service and the day it begins earning inside an approved tariff, months or years can pass, depending on the state and the procedural path. During that window the capital sits on the balance sheet but not in the rates.

The allowed return is a ceiling, not a promise

Here is the error that most sector research reproduces: the authorized return on equity is treated as though it were the return actually earned. It is not.

Work by Regulatory Research Associates, the regulatory service inside S&P Global, makes the movement unmistakable. In 2018, earned returns on equity at U.S. electric utilities averaged 105 basis points above authorized. By 2023 they were 84 basis points below. That is a swing of roughly 190 basis points in five years, on a metric that sits between nine and ten percent. In relative terms, about a fifth of the real return on shareholder capital moved — while the authorized number itself barely changed.

The cause is the same as the wedge: inflation, higher rates, and a spending pace running faster than the approval process. It is also why rate case filings surged, and why utilities routinely cite underearning in testimony as the reason for filing. Some of the largest rate cases ever seen in the United States were filed in exactly this period.

For an investor the implication is blunt. The authorized return is the ceiling of the room, not the floor. In periods of rising construction cost and slow procedure, the distance to that ceiling widens rather than narrows. Anyone building a valuation on the authorized figure is systematically modeling the optimistic case.

Germany prices the identical wire far lower

The most instructive part of the story appears when the same physical asset sits under a different regulator. For the fourth regulatory period — electricity 2024 through 2028, gas 2023 through 2027 — the Bundesnetzagentur, Germany’s federal network agency, set two equity rates in its decision of 20 October 2021: 5.07 percent pre-tax for new assets and 3.51 percent pre-tax for existing assets. In the third regulatory period the equivalent numbers were 6.91 and 5.12 percent.

Placed side by side, the scale of the difference becomes hard to ignore:

Parameter	United States (typical rate case)	Germany (4th regulatory period)
Allowed return on equity	about 9.5–9.7%	5.07% new assets / 3.51% existing assets, pre-tax
Recognized equity ratio	about 50%	40% (equity above that earns only the lower rate)
Duration of the determination	until the next rate case, often 2–4 years	five years, here 2024–2028
Interim adjustment	formula rates, riders, construction work in progress	capital cost surcharge with a variable base rate

The second row matters at least as much as the first and is almost never mentioned. Under section 7 of the German electricity network charges ordinance, operationally necessary equity earns the equity rate only up to 40 percent of operationally necessary assets. Anything above that threshold receives the considerably lower debt-linked rate. A German network operator that finances itself conservatively is penalized by the rulebook for its prudence.

Multiply the two effects and the recognized equity return on identical assets lands somewhere between a quarter and a third of the U.S. level. Same wire, same demand, same physics — a completely different income statement. That is not a market inefficiency. It is a policy decision, retaken every five years.

How contested that decision is can be read in the litigation. The Düsseldorf Higher Regional Court held the determination unlawful in a ruling dated 30 August 2023 and ordered the agency to set the rates again, faulting among other things the insufficient plausibility testing of the market risk premium. Germany’s Federal Court of Justice reversed that in December 2024 and sided with the regulator: the agency enjoys discretion in choosing its methodology, and a court may not substitute a methodology of its own. The industry association BDEW publicly called the judgment incomprehensible. In October 2025 the Düsseldorf court took up the equity return question again. Five years on, the question of what a wire is allowed to earn remains open.

The honest counter-argument: the capital cost surcharge

Quoting 5.07 percent in isolation is too easy, and the counter-argument belongs in this analysis. For investment made between cost audits, German regulation provides a capital cost surcharge. In a determination published in January 2024 the Bundesnetzagentur restructured how that surcharge is remunerated: an annually variable base rate derived from the yield on outstanding domestic bonds, plus a constant risk premium of three percent.

In practice that produced roughly 7.23 percent excluding, or 8.25 percent including, German trade tax for 2023, and subsequently 6.74 percent excluding and 7.69 percent including trade tax. Remuneration runs through a blended rate assuming a flat 40 percent equity and 60 percent debt structure.

That materially changes the picture. The marginal return on newly invested capital in Germany sits far closer to U.S. levels than the headline 5.07 percent suggests. What remains structurally low-yielding is the installed base — the 3.51 percent on existing assets. For an investor this is the distinction that matters: the existing European network estate is a low-yielding, bond-like asset; its growth is not. A company such as E.ON, which ended 2025 with a regulated asset base of roughly 48 billion euros and intends to invest another 48 billion euros through 2030 at around nine percent annual RAB growth, is deliberately shifting its mix toward better-remunerated new assets. Roughly 40 of those 48 billion euros are earmarked for networks, about three-quarters of it in Germany.

The other side of that trade sits in the same presentation. For 2026 the group guides to adjusted EBITDA of 9.4 to 9.6 billion euros and adjusted net income of 2.7 to 2.9 billion euros, with a dividend policy of annual increases of up to five percent through 2028. Nine percent asset growth, up to five percent dividend growth. The wedge here is wider than at Duke, not narrower.

The risk that appears in no investor deck

A transmission line is depreciated over 40 years or more. A data center lease runs 10 to 15 years. An AI training cluster is technologically obsolete in three to five. Those three clocks do not agree, and the difference eventually lands on somebody.

Regulators have noticed. The most consequential precedent so far is AEP Ohio’s data center tariff, adopted with minor adjustments by the Public Utilities Commission of Ohio on 9 July 2025 and effective 23 July 2025. Its core terms:

Term	Provision
Minimum take	at least 85% of contracted capacity, billed regardless of consumption
Ramp-up window	four years, with the minimum billing percentage rising over time
Exit fee	three years of minimum charges after a defined initial period
Applicability threshold	new or expanded projects above 25 MW, mobile data centers above 1 MW
Application fee	$10,000 to $100,000 depending on capacity

What is remarkable is less the content than the outcome. The commission rejected objections from major technology firms and industrial users and held that infrastructure costs must be borne by those who cause them rather than shifted onto general ratepayers. The stipulation was supported by commission staff, the state consumers’ counsel and several industrial energy groups.

For a shareholder the implication cuts both ways. In the near term, an 85 percent minimum take over a long contract is the best revenue quality this sector has ever seen — effectively a secured cash flow stream. In the long term, that same clause does not remove the risk; it only relocates where it surfaces. If an operator is no longer solvent in eight years, or simply pays the exit fee and walks, the line stays in rate base for another three decades. At that point a commission decides again whether the asset is still used and useful. If the answer is no, the term for it is stranded cost, and stranded cost generally lands on equity.

Investors should therefore read the quarterly disclosures for contract maturity rather than pipeline size. Southern Company reports more than 17 gigawatts of large-load contracts, including a 3.2 gigawatt agreement running 25 years, with roughly 8 gigawatts in late-stage development and a pipeline above 75 gigawatts. Duke Energy cites 7.8 gigawatts of signed electric service agreements against a 15.4 gigawatt project pipeline. The signed number is the one that binds; the pipeline is a statement of intent.

What this means for valuation

The opening formula implies a valuation rule that works far more cleanly for regulated utilities than any price-earnings multiple. If a business durably earns a return r on its book equity, grows at rate g, and investors demand a return k, then approximately:

Price-to-book ≈ (r − g) / (k − g)

The consequence is uncomfortable. A utility is worth more than its book value only when the allowed return exceeds its shareholders’ cost of capital. When it does not, every incremental dollar invested destroys value — even though it raises reported earnings in absolute terms. Growth in this sector is not automatically good.

Case	Return on book	Growth	Required return	Implied price-to-book
U.S. utility earning its authorized return	9.6%	5.0%	7.5%	1.84
U.S. utility 84 basis points short	8.8%	5.0%	7.5%	1.52
German installed network base	3.5%	1.0%	6.5%	0.45
German new network investment	7.7%	9.0%	6.5%	value-accretive, formula breaks down

The table is a simplification: it assumes constants, and it ignores taxes, unregulated segments and balance sheet effects. But it explains in one line what many investors experience as a puzzle — why U.S. utilities have traded at a clear premium to book for years while pure European network assets trade close to it. The difference is not sentiment. It is a number in an administrative decision.

The last row also shows the limit of the model. When growth exceeds the required return the denominator turns negative and the formula collapses. Mathematically that is trivial; substantively it is the whole European network story. A period of exceptional investment growth is by definition temporary, and a valuation must not extrapolate it into perpetuity.

Three scenarios to 2030

Scenario	Trigger	Effect on earnings	Early warning sign
Base case	capital plans execute, allowed returns hold	EPS growth 1–2 points below rate base growth	equity issuance in line with guidance, no cuts in rate cases
Political pushback	bills rise faster than wages, affordability becomes an election issue	allowed returns trimmed, costs reallocated to large customers	legislation capping returns, consumer advocates filing as complainants
Demand rolls over	AI capex cycle turns, data center projects cancelled	stranded cost disputes, write-downs against rate base	pipeline shrinking faster than signed contracts grow

The middle scenario is the underpriced one. The allowed return is not a law of nature but an administrative act, set in a political environment where household electricity bills are currently very visible. The German case shows the direction such pressure can take: from 6.91 percent in the third regulatory period to 5.07 percent in the fourth, at a moment of historically high investment need. The fifth period begins in 2028 for gas operators and 2029 for electricity; the regulator opened the consultation in January 2024 with a framework paper that contemplates, among other things, a flat WACC approach. For anyone holding a European network operator, that determination is the single most important event of the coming years — more important than any quarterly print.

A practical checklist

All of this reduces to a short but unusually discriminating checklist. First, put the guided rate base growth and the guided EPS growth from the same investor deck next to each other. The difference is the price you pay for the growth. Second, find the actual earned return on equity in the quarterly materials and compare it with the authorized figure. Several consecutive quarters well below it mean either a difficult regulatory jurisdiction or weak cost control, and both feed straight into the multiple. Third, check the ratio of signed large-load contracts to the headline pipeline. Fourth, look for announced equity issuance; it lives in the financing plan and rarely in the headline.

Two structural points are worth adding for anyone investing across borders. Utilities are among the most interest-rate-sensitive equities in the market, and not only through discounting: the allowed return itself follows the rate cycle with a lag, which means a rate shock hits the share price long before it reaches the tariff. That lag is precisely what produced the swing from 105 basis points of overearning to 84 basis points of underearning, and it will run in the other direction if policy rates fall and commissions are slow to reset returns downward.

The second point is currency and jurisdiction. A U.S. investor holding a European network operator is buying a euro-denominated, regulator-set cash flow whose allowed return is roughly half the domestic equivalent — a position that only works if the growth of the asset base does the heavy lifting. Withholding tax treatment differs by treaty and account type, and for an income-oriented position that difference compounds. None of this argues against the sector. It argues against filing it under “defensive” simply because the underlying business is.

What remains

The AI boom is real for utilities, and it is large. A planned 1.4 trillion dollars of investment over five years, more than 55 gigawatts of large load already connected, contracts running 25 years — these are not projections, they are decisions. What that wave does not do is convert one-for-one into shareholder return, and the reason sits not in an annual report but in a regulatory order.

The decisive question at a regulated utility is never how much electricity gets sold. It is: how much capital is allowed into rate base, at what rate is it remunerated, how much of it must be funded with new shares, and how long before the tariff catches up with reality. Line those four numbers up and Duke Energy shows seven to nine percent asset growth against five to seven percent earnings growth; Southern Company nine against seven to eight; E.ON nine against up to five. The gap is the same order of magnitude everywhere, because the same mechanics operate everywhere.

None of that makes utilities bad investments. It makes them investments whose return can be calculated from four parameters rather than inferred from a story. And it moves the question from how fast the data centers are growing to who ultimately decides what the copper in the ground is allowed to earn. That person does not work on an exchange.