The Quiet Transformation of Your Power Bill
Somewhere in rural Virginia, a squat, windowless building the size of several football fields hums with the sound of cooling fans and the faint crackle of electrical current. It draws more power than a small city. Down the road, a household opens its monthly electricity statement and wonders why the bill keeps climbing.
Whether those two things are connected is the heart of a growing political fight.
The artificial intelligence infrastructure boom has triggered one of the largest surges in electricity demand the United States has seen in decades. Data centers, the physical backbone of AI model training and inference, are being constructed at a pace that has caught grid operators and regulators off guard. The question of who pays for the transmission lines, substation upgrades, and generation capacity required to serve these facilities has become one of the most consequential and least-discussed policy fights in American energy politics.
Critics argue the answer, too often, is ordinary ratepayers.
The Scale of the Demand Shock
To understand the stakes, you have to grasp just how much electricity AI data centers actually consume. A traditional internet data center might draw 20 to 50 megawatts. A hyperscale AI training campus, the kind Microsoft, Google, Amazon, and Meta are now constructing across the country, can demand 500 megawatts or more from the grid. Some proposed campuses exceed one gigawatt, enough to power roughly 750,000 average American homes.
The Electric Power Research Institute has projected that data centers could account for between roughly 4.6 and 9.1 percent of U.S. electricity consumption by 2030, depending on the growth scenario. Goldman Sachs Research estimated that global data center power demand could grow by 160 percent by 2030. Grid operators from PJM Interconnection, which oversees the mid-Atlantic and Midwest regions, to the Electric Reliability Council of Texas have issued warnings about the strain on existing infrastructure.
In Northern Virginia, the self-proclaimed “Data Center Alley” and arguably the densest concentration of computing infrastructure on Earth, Dominion Energy has repeatedly revised its load forecasts upward. The utility has submitted integrated resource plans projecting substantial new generation capacity, driven in significant part by data center demand. Under traditional utility cost-recovery models, the utility’s customers, not Amazon or Google, are the default backstop for infrastructure investment under traditional utility cost-recovery models.
How Utility Regulation Became a Battlefield
American electricity markets are governed by a patchwork of state public utility commissions and federal regulators at the Federal Energy Regulatory Commission (FERC). The basic model, built over a century of regulatory precedent, works as follows: when a utility needs to upgrade its infrastructure, it files a rate case, regulators review the prudency of the investment, and if approved, costs are spread across the utility’s customer base through rates.
This model was designed for a world where demand grew slowly and predictably. It was not designed for a world where a single hyperscale customer might require as much new infrastructure as an entire mid-sized metropolitan area.
The critical fault line is what utility lawyers call “cost causation.” The principle holds, reasonably enough, that those who cause grid costs should bear them. Applied to data centers, it would suggest that tech companies, not the general ratepayer population, should pay for the dedicated transmission and distribution upgrades their facilities require. But in practice, critics say, regulators in many places have historically spread these costs across all customers.
That is beginning to change, though inconsistently and slowly. In November 2025, Virginia’s State Corporation Commission approved a new rate class for Dominion customers using more than 25 megawatts, with long-term contracts and minimum demand charges intended to protect other ratepayers. In 2024, FERC issued Order 1920, the most sweeping transmission planning reform in roughly a decade, which addressed some long-range planning requirements. But the order did not fully resolve how costs for new large-load interconnections should be allocated. State commissions have moved at varying speeds and with varying philosophies.
Virginia’s State Corporation Commission has been a focal point as it reviews Dominion Energy’s rate cases and capital spending plans, with consumer advocates arguing that ordinary households risk being asked to subsidize infrastructure that primarily benefits large corporations. Their concern is an asymmetry: data center operators benefit from grid access, yet, advocates argue, their interconnection fees and direct cost assignments can represent a small fraction of the infrastructure cost their facilities impose.
Georgia Power, serving a state that has attracted significant data center investment partly through aggressive tax incentives, has faced similar scrutiny from the Georgia Public Service Commission over its capital expenditure plans. Similar debates are under way in Texas, Ohio, Nevada, and other states where tech companies have clustered.
The Tech Industry’s Counterargument
The technology industry does not accept this framing quietly. Trade associations and individual companies argue, with some legitimacy, that data centers bring substantial economic benefits to host communities: jobs, tax revenues, and in some cases direct investment in grid modernization that ultimately benefits all users.
Microsoft, Google, and others frequently point to their investments in renewable energy procurement, battery storage agreements, and in some cases direct funding for new generation projects. Amazon Web Services has positioned itself as one of the largest corporate buyers of renewable energy globally, pointing to power purchase agreements for solar and wind that add capacity to regional grids.
The industry also argues that new large industrial loads, including data centers, effectively help utilities spread fixed infrastructure costs across a larger revenue base, which can theoretically reduce per-unit costs for other customers. This is the classic argument for load growth as a ratepayer benefit, and it is not without merit under certain conditions.
Some consumer-side energy economists argue that the benefit-spreading argument obscures more than it reveals. When new load requires large capital investments in dedicated infrastructure, the math often runs in the opposite direction: the new load drives new costs faster than it generates new revenue to offset them, at least in the near term. The timing mismatch matters enormously for ratepayers who bear carrying costs through their bills while waiting for promised benefits to materialize.
There is also the question of what happens when a data center shuts down or relocates. Unlike a new neighborhood of homes, large industrial customers can and do leave. If a hyperscale facility closes and the transmission lines built to serve it are “stranded,” those costs remain on the rate base, paid by whoever is left.
The Locational Politics of Data Center Placement
Data center siting decisions are not random. They reflect a competitive marketplace in which states and localities bid against one another using tax abatements, streamlined permitting, and favorable utility rate structures. Virginia has offered substantial data center tax exemptions. Texas has historically offered low electricity costs through its deregulated ERCOT market. Other states have sweetened deals with workforce development incentives and expedited grid interconnection.
Critics say this competitive dynamic risks a race to the bottom on cost accountability. States that impose stricter cost-causation rules on large industrial customers could lose data center investment to jurisdictions with more permissive regulatory environments. The worry is a collective action problem: each state has an incentive to ask less of tech companies, even if the aggregate effect is to shift infrastructure expenses onto residential and small commercial ratepayers.
Advocates note that some communities courting data centers have significant populations of low-income households, for whom electricity cost increases are not a minor inconvenience but a genuine hardship.
Environmental justice advocates have added another dimension to the critique. They argue that data centers tend to cluster in areas with existing transmission capacity, which often means they are located near legacy fossil fuel generation infrastructure. The heavy cooling loads they impose can increase peak demand, which grid operators frequently meet with the most carbon-intensive “peaker” plants. Nearby communities, often lower-income and disproportionately communities of color, bear the air quality consequences.
What Regulators Are Actually Doing About It
The regulatory response has been uneven but not entirely toothless. A handful of state commissions have begun requiring that large new industrial customers pay a greater share of the direct infrastructure costs attributable to their interconnection. Some have implemented “large customer” tariff structures that require deposits or upfront contributions toward grid upgrades before service begins.
In Nevada, where data center development has accelerated near Reno and Las Vegas, the Public Utilities Commission has examined rate design questions around large commercial customers. In Texas, the deregulated market structure means that transmission and distribution costs are handled differently, with large industrial customers sometimes negotiating directly with transmission providers, though the cost socialization problem persists in subtler forms.
FERC’s broader transmission reform efforts under Order 1920 have tried to address long-range planning, including 20-year assessments that account for anticipated load growth. Critics argue this is necessary but insufficient without cleaner rules on who pays for what when a specific large customer triggers specific infrastructure needs.
Some observers have pointed to the “beneficiary pays” principle as the most defensible framework going forward. Under this approach, the entity whose load growth drives a required grid investment would be required to fund that investment, potentially with a clawback mechanism if the load does not materialize as projected. In late 2025, at the Department of Energy’s direction, FERC opened a rulemaking proceeding on large-load interconnection, built on a DOE proposal that would have large customers bear the network upgrade costs they cause, and on June 18, 2026, rather than issuing a single final rule, it issued show cause orders to the six regional grid operators it regulates, directing them to justify or reform their large-load interconnection rules, including on cost allocation.
The challenge is technical as well as political. Attributing grid costs to specific customers with precision is genuinely difficult. Transmission systems are networks; an upgrade in one location may relieve constraints across a wide area, making it hard to say that any single customer “caused” the investment. That ambiguity is contested in regulatory proceedings.
Where the Money Goes, and What Comes Next
The financial stakes are enormous and growing. Some estimates of the U.S. transmission investment needed over the coming decade run into the hundreds of billions of dollars. Even a modest shift in cost allocation, moving from full socialization to a 50 percent large-customer cost assignment, for example, would represent tens of billions of dollars redirected away from ordinary ratepayers.
For the technology companies involved, these are not existential sums. Microsoft reported revenue of roughly $245 billion in its fiscal year 2024. The capital these companies are deploying on data center construction is very large. Supporters of stricter cost assignment argue that infrastructure contributions that would materially protect ratepayers are small relative to those budgets.
For a retired schoolteacher, a factory worker, or a small restaurant owner, upward pressure on electricity rates that they attribute to corporate and regulatory decisions is something quite different.
Consumer advocates and some elected officials have been raising these issues more publicly, treating data center cost allocation not merely as a technical rate-design question but as a matter of economic fairness and democratic accountability. The framing is shifting: this is not just about electrons and substations. It is about who in American society bears the costs of the infrastructure that makes artificial intelligence possible.
The outcome of these regulatory fights over the next few years will shape electricity bills for tens of millions of Americans for decades. Grid infrastructure, once built, tends to stay in rate bases for 30 to 40 years. Decisions made in state capitals and FERC proceedings between now and 2030 will lock in cost allocation frameworks with extraordinary staying power.
The AI boom is, by most accounts, still in its early innings. The data centers being announced today will draw power for years and will likely be followed by still larger facilities as AI systems continue to scale. If regulators do not act to establish clear, durable, and fair cost-causation principles now, critics warn the default outcome is that costs will land on whoever is least organized, least politically connected, and most reliant on electricity as a basic necessity.
In the critics’ view, that is the rest of us.