The bull case for regulated utilities is unusually simple. A utility earns an authorized return on the capital it invests in its system. More investment means a larger rate base, which means higher earnings. Data centers are creating demand that requires enormous investment in generation, transmission, and substations. Therefore: buy utilities.
Every step of that is correct. The step that gets skipped is that rate base growth requires permission, and the body granting permission is politically accountable.
The scale of what is pending
Utilities requested a record $31 billion in rate hikes in 2025, more than twice the near-record set in 2024, according to the nonprofit PowerLines. At least 210 utilities raised or proposed raising bills last year, with a potential cumulative effect of $67 billion by 2028.
Residential electricity prices rose 7% in 2025 alone. Some regions have seen far more: between March 2021 and March 2026, average residential rates rose 94% in Washington, D.C., 74% in Maryland, 73% in Maine, and 58% in New York.
The timing detail is what makes this an active rather than resolved risk. Nearly half of the 2025 requests were still pending entering 2026, and as PowerLines' executive director put it, many of these increases have not actually hit people's wallets yet. The political reaction is therefore still building. It is ahead of the market, not behind it.
What Wall Street has already noticed
The analyst framing is the most useful thing in this story. Jefferies' power and utilities analyst Julien Dumoulin-Smith described the industry's 2026 narrative as shifting from "capex growth at all costs" to capex growth with a customer permission slip, adding that it is no longer enough for utilities to say they care about affordability, because regulators and investors will demand proof of proactive behavior.
That is a precise description of a changed regulatory environment. It does not say the capex is not coming. It says approval is no longer close to automatic, which changes both the timing and the certainty of the earnings that capex is supposed to produce.
Morgan Stanley's utilities analyst has raised the same concern from the other direction, flagging electricity affordability as a constraint on utility earnings growth. When two sell-side shops covering the sector both name affordability as the binding issue, that is not a fringe worry.
The attribution fight, and why it matters less than it seems
The evidence on whether data centers are actually driving residential increases is genuinely contested, and it deserves to be presented that way.
The Electric Power Research Institute found that data centers put downward pressure on average electricity prices through 2024, on the logic that a large customer paying into a grid built for peak demand spreads fixed costs across more consumption. A study by Energy + Environmental Economics, commissioned in connection with the data center industry, found no historical evidence that data centers drove residential cost increases under existing rate structures, attributing rises instead to inflation in labor, materials, and financing. Lawrence Berkeley National Laboratory has pointed to equipment costs, an aging grid, and clean energy requirements as contributors.
Against that, other analyses attribute roughly $23 billion in customer price increases running through 2028 primarily to expected data center demand.
Here is why the resolution matters less than it appears for anyone assessing the sector. Regulatory outcomes follow political salience, not attribution studies. A voter opening a bill that jumped from $100 to $281 is not weighing the relative contributions of inflation, transmission depreciation, and interconnection queues. Lawmakers in more than 30 states have introduced over 300 data-center-related bills this year, and a bipartisan push has urged state commissions to reject proposals that raise residential rates to serve data center load.
Whether or not data centers cause the increases, they are the identifiable cause, and identifiable causes are what regulation responds to.
The reversal already underway
The concrete policy response is a structural change that runs against decades of practice.
Regulators are increasingly setting a separate rate class for data centers so those customers bear the costs of the new transmission lines and substations required to serve them, rather than spreading those costs across all ratepayers. Historically, large industrial customers received lower rates than households, as an economic development incentive. At least 36 states offer tax incentives to attract data centers.
Large load tariffs invert that. And the inversion has direct consequences for the investment case, because a data center required to pay the full incremental cost of serving it is a data center with weaker economics, which affects how many get built and where.
The negotiating dynamic is getting sharp. When DTE Energy offered to freeze residential rates for two years if a large data center came online as planned, Michigan's attorney general likened the offer to a ransom note. That reaction, from a state's chief legal officer, is a reasonable proxy for where the politics sit.
The demand number that explains the anxiety
One figure conveys why regulators are being cautious about approving capital plans built on projected load. Utilities received data center interconnection requests totaling at least 700 gigawatts in 2025, exceeding the 477 gigawatts of electricity the entire United States consumed in 2023.
Most of those projects will never be built. Interconnection requests are cheap to file and developers routinely queue the same project in multiple territories. But utilities plan and spend against them, and that creates the stranded asset problem regulators are now explicitly worried about: capital committed to serve demand that may not materialize, with existing ratepayers holding the bill.
For an investor, that cuts both ways. Stranded assets are a risk to utility earnings if regulators disallow recovery. They are a risk to ratepayers if regulators allow it. Which way a commission resolves that is a political question, and the political environment is tightening.
How to read the trade
None of this makes the AI power thesis wrong. Electricity demand is genuinely growing after two decades of flat load, that growth genuinely requires investment, and utilities genuinely earn returns on investment. The structural case is real.
The adjustment is to stop treating regulatory approval as a formality in the model. The sector's earnings growth depends on commissions granting rate increases at a scale and pace that is now politically contested, in a year when nearly half of a record request backlog is still pending and voters are receiving bills that have not yet fully reflected what was already approved.
Dumoulin-Smith's phrase is the one to keep. Capex growth with a customer permission slip. The permission slip is the variable, it is not priced with much precision, and unlike commodity prices or interest rates, it is decided by people who face election.