The artificial intelligence revolution has a power problem. And millions of American households could end up paying to solve it.
Across the country, technology companies are racing to construct enormous data centers, facilities packed with specialized computer chips that consume staggering quantities of electricity. Microsoft, Amazon, Google and Meta are investing heavily in computing infrastructure. Their ambitions are reshaping the electricity industry, accelerating demand for power plants, transmission lines and substations.
But a less visible financial question is emerging inside utility regulatory proceedings: if an electric company builds billions of dollars in infrastructure to serve a new data center, who ultimately pays? The technology company, utility shareholders, or families and small businesses already connected to the grid?
The Charlotte electricity bill
Duke Energy Carolinas initially sought an approximately 18% increase in residential electricity rates. A proposed settlement reduced the average residential increase to approximately 9.5%. North Carolina Attorney General Jeff Jackson refused to support the agreement, arguing that the increase remained excessive. The settlement was proposed, not a final approved rate.
Duke’s regulatory disclosures describe approximately $496 million in cumulative additional retail revenue under its proposed two-year settlement. That is a figure for the broader rate proposal, not an independently established cost of AI infrastructure.
The number that changes the conversation
According to the North Carolina Department of Justice, more than 80% of Duke’s projected new electricity demand comes from data centers and other large electricity users. This does not mean those customers account for 80% of current consumption or 80% of the proposed residential rate increase. It does mean large customers are expected to drive much of the next wave of demand.
A hyperscale data center can require hundreds of megawatts of electricity. Meeting that demand may require generating capacity, upgraded transmission networks and dedicated substations. Those investments are expensive and long-lived, even if the customer whose forecast justified them changes plans.
A business model hiding in plain sight
Regulated utilities generally recover prudent operating expenses and earn an authorized return on approved capital investments. Duke’s proposed settlement includes a 9.8% authorized return on equity. That is not a guaranteed profit on every dollar spent, but infrastructure expansion can expand the investment base on which utility shareholders are permitted to earn returns.
The interests of technology companies, utilities and existing customers can align: large new customers may bring substantial revenue and help spread fixed costs. But if investment costs are disproportionately recovered from others, or large customers depart before paying their share, the benefits and risks can become unevenly distributed.
The gap in North Carolina’s proposed protections
In October, Duke reached a proposed agreement on a large-load tariff for exceptionally large electricity customers. Qualifying customers would pay for at least 75% of contracted demand each month and pay upfront for certain facilities constructed exclusively for them, including dedicated substations.
These are meaningful safeguards against speculative projects. Yet the North Carolina Department of Justice noted that the proposal does not explicitly require large customers to cover the full costs of new electricity generation attributable to their demand. A developer might fund its dedicated connection while shared power plants and transmission networks still create costs across the broader utility system.
The proposed rules also do not establish an entirely separate rate class for data centers, and some projects with earlier contracts may fall outside the new provisions. The question is whether the guarantees cover the full incremental costs and risks.
Austin offers a revealing comparison
Austin’s adopted fiscal year 2027 budget estimates that a typical taxpayer and utility customer will pay approximately $285.84 more annually across property taxes, utility charges and other fees. Austin Energy has identified debt service, compensation, transmission expenses and infrastructure requirements among its financial pressures.
Austin Energy has policies requiring large customers to cover 100% of applicable line-extension and connection infrastructure costs, subject to specified exceptions. That addresses direct connections but does not automatically resolve the costs of additional generation or shared transmission.
As reported by the Texas Tribune in August, Austin Energy was serving existing data centers but had no large-scale AI hyperscale data centers operating within the city at that time. The evidence therefore does not establish that Austin’s current household increases are primarily caused by hyperscale AI development.
Texas’s speculative demand pipeline
In June 2026, ERCOT reported more than 438,000 megawatts of large-load connection requests, nearly 89% associated with data centers. These are potential connection requests, not proof of completed projects or committed operating demand. Some proposals may be speculative, delayed or overlapping.
The planning dilemma is substantial. Build too slowly and electricity shortages may constrain growth. Build too aggressively and customers may be left paying for infrastructure that exceeds actual demand.
The risk nobody wants to own
Imagine a developer requesting enormous electricity capacity, prompting a utility to begin investing in equipment with long lead times. If the developer later cancels, the infrastructure investments may already be underway. A well-designed contract would require deposits, minimum payments, termination provisions and other guarantees sufficient to cover costs attributable to the project. A weaker arrangement could expose other customers.
A 75% minimum-payment commitment can offer substantial protection, but its effectiveness depends on contract duration, actual infrastructure costs, creditworthiness and early-termination provisions.
Federal regulators are worried, too
In October, the Federal Energy Regulatory Commission accepted additional Duke transmission-related protections, including revenue commitments based on at least 75% of planned demand over 15 years. Commissioners nevertheless highlighted concerns about stranded costs and potential effects on existing customers. These federal transmission rules are distinct from North Carolina’s proposed retail tariff.
What Duke Energy says
Duke argues that its investments are needed to maintain reliability and accommodate growth. It emphasizes upfront payments and financial commitments intended to protect existing ratepayers. Large customers could also bring enough revenue to benefit the wider system. Those arguments deserve examination alongside the concerns raised by consumer advocates.
Not every dollar of an electricity rate increase is attributable to data centers. Aging infrastructure, reliability, financing and broader growth also contribute. Determining any AI-related subsidy requires project-level cost accounting, not simply comparing data-center construction with rising household bills.
The investigation that still needs to happen
The available evidence establishes a meaningful risk of cost shifting, but does not establish exactly how much North Carolina households are paying specifically for AI-related infrastructure. Regulators and utilities should disclose which investments are driven by data-center demand, what guarantees those customers provide, who pays if projects fail, and how costs are allocated to residential ratepayers.
The most revealing question is this: how much would an average Charlotte household’s bill change if projected data-center expansion were excluded from Duke’s investment plans? A transparent counterfactual analysis could separate AI-driven expansion costs from investments that would be needed regardless.
The next industrial revolution needs a fair financial foundation
Artificial intelligence may transform medicine, science, manufacturing and education. Building infrastructure to support it is not inherently wrong. Nor is allowing utilities to earn reasonable returns on prudent investments. The problem arises when the risks of private expansion are distributed in ways ordinary customers cannot see or challenge.
Charlotte residents deserve clear answers about which investments are driving their bills, who benefits and who is contractually obligated to pay. Technology companies can pursue the opportunities created by AI. Utilities must provide reliable service. Regulators must protect the public interest.
If Big Tech needs the power, Big Tech should pay the costs directly attributable to providing it. That is not opposition to innovation. It is a prerequisite for making innovation economically sustainable.
Sources and methodology
NC Department of Justice: proposed large-load tariff; NC Department of Justice: rate increase dispute; WFAE: tariff and unresolved costs; FERC: transmission protections; Duke Energy: proposed settlement; City of Austin budget.
Editorial disclosure: This is a research-based explanatory investigation drawing on public records and published reporting available through October 25, 2026. No original interviews were conducted. A specific AI-attributable portion of the proposed residential rate increase has not been independently established.