Why Your Electricity Bill Is Going Up Because of the Data Center Down the Road

Intended for ratepayers, policymakers, and energy market participants trying to understand who actually pays for the infrastructure a new data center requires — and why that answer has been changing state by state since 2025.

Executive Summary

Rising electricity bills are increasingly tied to data center load growth, but not because data centers bill residential customers directly. The mechanism runs through the utility rate case: when a utility needs new generation, transmission, or distribution capacity to serve a large new customer, a state public utility commission (PUC) decides who pays for it. Absent a specific rate class for that new customer, those costs have historically been spread across the entire ratepayer base — including residents who think they see none of the benefit. That default is now being challenged. As of June 2026, 24 states have approved at least one large-load tariff designed to make data centers cover their own infrastructure costs, with Dominion Energy's Virginia GS-5 tariff, Oregon's Schedule 96, Texas's SB 6, and Ohio's AEP model among the most developed examples. But state-level tariffs only address part of the bill. PJM's own market monitor documented $13 billion in costs already distributed across PJM-region ratepayers from data center load additions in the first quarter of 2026 alone — costs that move through a separate, federal layer of cost allocation that most state tariffs don't touch. A third channel sits outside any tariff structure entirely: new gas-fired generation built to serve data center load buys fuel on the same market as every existing gas plant, and EIA's own modeling shows that added demand raising natural gas prices system-wide — a cost that shows up on electricity bills with no data center, and no large-load tariff, anywhere in the chain.

A note on framing before the mechanics: this is not an argument against new data centers. They are necessary infrastructure, and their buildout has historically moved about as slowly as nuclear power plant buildout — measured in years per project, not months. What changed is not the existence of data centers. It is the pace of demand. AI tools drove a sudden, largely unplanned-for surge in load on top of that already slow-moving base, and it is that demand shock, not data centers themselves, that has forced the rate-case and cost-allocation fights described below. Most readers of this piece are also, whether they realize it or not, already data center users — see the FAQ below.

How a Rate Case Actually Works: Who Decides Who Pays

A utility does not set its own rates. When it needs to recover the cost of new generation, transmission lines, or substation upgrades, it files a rate case with its state public utility commission or public service commission (PUC/PSC), which reviews the request and decides how those costs get allocated across customer classes — residential, commercial, industrial, and, increasingly, a separate class for very large loads like data centers.

The default setting, absent a specific large-load tariff, has generally been to treat new infrastructure investment as a shared system cost: rate-based and recovered from the entire customer base through general rates, on the theory that grid investment benefits everyone over time. That assumption works reasonably well when new load grows gradually and predictably. It works far less well when a single data center campus adds hundreds of megawatts of demand in a few years — infrastructure that exists because of one customer, paid for by all of them.

The Fight Over the Default: Who Should Absorb the Risk

This is not a settled question, and it is being actively litigated in front of regulators right now. In California, Pacific Gas & Electric's proposed large-load tariff before the California Public Utilities Commission has become a flashpoint precisely over this issue: PG&E's structure would recover system upgrade costs through general rates rather than assigning them directly to the large-load customers driving the need for them, on the argument that this mirrors how the utility recovers cost for any other capital expenditure. Ratepayer advocacy groups, including TURN and the Public Advocates Office (CalAdvocates), have pushed back, arguing that infrastructure clearly driven by data center additions should be assigned to those customers directly rather than spread across the general transmission rate base — where the risk lands on all ratepayers if a data center's projected demand doesn't materialize.

That risk is not hypothetical. It is the core design problem regulators are now trying to solve: protect ratepayers from being left holding stranded infrastructure costs if a speculative data center project shrinks or doesn't proceed, while still giving legitimate projects a predictable cost structure to plan around.

The Fix Utilities Are Adopting: Large-Load Tariffs

A large-load tariff is a separate, legally binding rate class specifically for very large customers — typically defined by a minimum demand threshold, often 20 to 25 megawatts or more — that determines how much they pay and under what conditions they connect to the grid. The key design principle across most of these tariffs is the same: require the large-load customer to cover the incremental infrastructure cost their addition requires, rather than letting that cost land on the general rate base.

Several state-level examples now anchor this trend:

Virginia — Dominion Energy's GS-5 rate class, approved by the State Corporation Commission in November 2025, is the most comprehensive AI-specific utility tariff enacted to date. It establishes a 14-year contract term for large loads above 25 megawatts, requires demand guarantees covering generation and transmission costs, and creates a separate cost allocation mechanism intended to protect residential and small commercial customers from absorbing costs tied to data center load additions.

Ohio — AEP Ohio's model requires data centers to cover 85 percent of their requested demand in minimum monthly charges, regardless of actual usage, which eliminates the risk of a developer securing grid capacity and then failing to deploy the load, leaving the utility with stranded infrastructure investment. This structure has become a reference point other utilities are adapting for their own large-load tariff proposals.

Oregon — Portland General Electric's Schedule 96, approved under the state's 2025 POWER Act and effective June 10, 2026, establishes a dedicated rate class for large loads above 20 megawatts, requires those customers to cover 100 percent of the distribution network upgrades their project requires, and sets minimum generation and transmission demand charges at 90 percent of contracted capacity. Contract terms scale with project size, from 10 years up to 30 years for loads of 220 megawatts or more.

Texas — SB 6, enacted in June 2025, addresses large-load growth on four fronts: ensuring transmission costs are properly allocated, establishing grid reliability protections, requiring transparent and credible load forecasting, and — notably for reliability — requiring large loads to share the load-shed obligation during periods of grid shortage, rather than being shielded from curtailment while residential customers absorb it. That last point matters directly for Texas: ERCOT's own emergency dispatch priority already places industrial processes last, behind hospitals and residences. SB 6 extends that same logic to who bears the cost and reliability burden of the infrastructure buildout itself, not just who gets curtailed first in an emergency.

The scale of this shift is significant. As of June 2026, 24 states have approved at least one large-load tariff and another 4 have proposals pending. A separate industry tracker puts the count higher when proposed tariffs are included: 51 approved and 26 proposed large-load tariffs or service rules across 36 states, targeting 60 different utilities, as of a March 2026 update.

The Piece No Large-Load Tariff Touches: What New Gas Demand Does to Everyone Else's Fuel Bill

Even a well-designed large-load tariff, one that makes a data center cover 100 percent of its own interconnection and demand costs, does not address a separate channel entirely: the fuel market.

Natural gas accounted for 40 percent of total U.S. electricity generation in 2025, and it is the fuel utilities are turning to fastest to meet new data center demand. When a new gas-fired plant is built to serve that load, it buys natural gas on the same regional and national market as every other existing gas-fired power plant. Added demand for that fuel raises its price system-wide — not just for the new plant drawing it.

This is not a hypothetical. The EIA's own February 2026 Short-Term Energy Outlook modeled a high data-center-demand-growth scenario and assumed an increase of roughly $0.50 per million British thermal units (MMBtu) in the cost of natural gas, driven specifically by the additional fuel demand required to meet incremental electricity load through increased utilization of gas-fired plants.

The pressure is already visible on the supply side. The capital cost to build a new combined-cycle gas turbine plant rose from under $1,500 per kilowatt of capacity in 2023 to $2,157 per kilowatt in 2025, and construction timelines have stretched roughly 23 percent longer, developments industry reporting has tied directly to data center-driven demand for new gas generation. Pipeline capacity, not just production capacity, is a live near-term constraint: during a January 2026 winter storm, national natural gas demand rose 20 percent week over week and Henry Hub spot prices spiked from an average of roughly $7.75 per MMBtu to $30 per MMBtu — an event that reflected pipeline throughput limits as much as production limits. With roughly 150 gigawatts of data center capacity currently in planning nationally against an existing 130 to 230 gigawatts of gas demand for power generation, a full build-out at that pace represents a 50 percent or greater increase in gas demand for electricity generation alone.

The mechanism that turns this into a bill increase for someone with no connection to any data center is a function of how wholesale electricity markets actually work, not a story about individual generators and their customers. Generators do not sell to specific customers the way a retailer does. Because electricity must be produced and consumed in the same instant, generators bid power into the regional grid operator's wholesale market, and the grid operator — PJM, across much of the mid-Atlantic and Midwest — dispatches generators in order from cheapest bid to most expensive to meet real-time demand. In most of these markets, every dispatched generator is paid the same clearing price for that interval, set by the bid of the last, most expensive unit needed to meet demand: the marginal unit.

Natural gas plants routinely set that marginal price. When gas fuel costs rise, the marginal gas plant's bid rises, and the higher clearing price it sets gets paid to every generator dispatched alongside it in that interval — a nuclear plant, a coal plant, a wind farm with no data center anywhere near it — not just to gas plants, and not just to the newest one predominately serving a data center. That higher clearing price is what utilities and load-serving entities pay for the power they buy off the market, and it flows through from there to the retail rates every connected customer pays. A large-load tariff can make a data center pay for its own poles and wires. It cannot change how the marginal price in a wholesale market gets set, or stop that price from applying to every generator dispatched at that moment.

The Gap Between the State-Level Fix and the Bill You're Already Paying

Here is the part that gets missed in most coverage of this issue: a well-designed state large-load tariff does not necessarily touch the layer of cost allocation where the largest dollar figures are currently showing up.

Cost allocation for a data center's power happens across at least three distinct regulatory layers: the Federal Energy Regulatory Commission (FERC) wholesale capacity and energy market layer, the regional grid operator's tariff layer governing transmission and interconnection (PJM, which originally stood for Pennsylvania, New Jersey, and Maryland, now covering much of the mid-Atlantic and Midwest), and the state PUC retail layer governing distribution service and local generation costs. State large-load tariffs, including all of the examples above, operate at that third layer. They do not directly govern the first two.

That distinction has real consequences. PJM's own market monitor, Monitoring Analytics, documented $13 billion in added costs distributed across PJM-region ratepayers from data center load additions in its first-quarter 2026 state of the market report — a figure driven by PJM's capacity auction mechanics under FERC-approved tariff rules, not by any state PUC rate case. A ratepayer in a PJM state with a newly approved large-load tariff may still see bill increases traceable to this federal layer, entirely separate from whatever protection the state tariff provides at the retail level.

Rural Community Economics: Where This Lands Hardest

The infrastructure buildout required to serve new data center load doesn't distribute evenly, and neither do its benefits. A useful comparison: an existing 1 gigawatt nuclear plant supports roughly 600 permanent jobs at $100,000 to $120,000 in annual compensation, versus roughly 30 jobs for a comparably sized gas plant. Currently, natural gas turbines carry a five- to seven-year fabrication backlog compared to advanced nuclear projects estimated to carry an eight- to twelve-year timeline, which means the faster-to-build generation option to meet near-term data center demand is frequently not the option that leaves a host community with a lasting employment base.

That creates a specific version of the cost-allocation problem for rural and host communities: they may absorb both a share of the infrastructure cost, if their state has not yet adopted a large-load tariff with real teeth, and the comparatively thin permanent job base that comes with faster-to-build generation technology — while the data center itself is not a large source of permanent local jobs either. Hyperscale facilities are highly automated and typically operate with as few as 20 to 30 permanent staff per 100 megawatts, according to a November 2025 industry workforce forecast. Virginia's own Department of Taxation reported the data center industry added just 1,610 jobs statewide in fiscal year 2025, while the industry received $1.9 billion in state tax benefits over that period — roughly $1.2 million in incentives per job created. Brookings researchers who compared employment outcomes across 93 counties and 770 data center facilities found that job creation claims made when these projects are pitched to communities were regularly overstated by a factor of three. A large-load tariff addresses the cost side of the equation host communities face. It does not address the jobs side, and on the jobs side, neither the faster-to-build generation technology nor the data center campus it serves brings much permanent employment to the table.

What Small Teams Should Do Now

• Before assuming a state's new large-load tariff has "fixed" rising bills, check which regulatory layer the actual cost increase is coming from — state PUC retail rates, or the PJM/FERC wholesale and capacity layer, which most state tariffs don't reach.

• If you are developing a reactor project intended to serve a data center customer, understand which large-load tariff regime the utility operates under before finalizing project economics — a Dominion-style demand guarantee, an AEP-style 85 percent minimum-take structure, and an Oregon-style 100 percent customer-funded distribution model each carry materially different cost profiles.

• If community relations is part of your siting strategy, be direct with host communities about the jobs case your specific technology supports. A gas-fired interconnection and a nuclear plant are not equivalent employment stories, even at the same megawatt scale.

• Track your own state's large-load tariff status. With 24 states approved and dozens more considering proposals as of mid-2026, the regulatory environment for who pays is still actively being written.

• Don't assume a large-load tariff means data center-driven bill increases have stopped. Fuel-price pass-through from new gas demand operates entirely outside tariff structures and affects every gas-fired generator's marginal cost, tariff or no tariff.

  
  

Frequently Asked Questions

Q: Why is my electricity bill going up because of data centers, if I don't use one?

A: When a utility needs new generation, transmission, or distribution infrastructure to serve a large new customer like a data center, a state utility commission decides how to allocate that cost. Absent a specific large-load tariff assigning those costs to the data center, they have historically been spread across the general rate base, including residential customers.

Q: I don't use AI tools. Doesn't that mean I'm not driving data center demand?

A: Not necessarily. Data centers host far more than AI workloads — email, cloud storage, online banking, streaming video, social media, and most modern software and websites run through them. Someone who never opens an AI chatbot is still very likely using a data center dozens of times a day without realizing it. The recent surge in new capacity and the large-load tariff fights described in this piece are driven specifically by AI-scale demand growth, but the underlying reliance on data center infrastructure has been part of ordinary internet use for over two decades.

Q: What is a "large-load tariff" and how does it work?

A: It is a separate, legally binding rate class for very large electricity customers, typically defined by a minimum demand threshold around 20 to 25 megawatts, that requires the customer to cover the incremental infrastructure costs their connection requires, rather than spreading those costs across all ratepayers.

Q: Has any state actually implemented this, or is it still theoretical?

A: It is implemented and active in multiple states. Virginia's Dominion Energy GS-5 tariff was approved by the State Corporation Commission in November 2025. Oregon's PGE Schedule 96 took effect June 10, 2026. Texas's SB 6 was enacted in June 2025. As of June 2026, 24 states have approved at least one large-load tariff.

Q: Does a state large-load tariff mean my electricity bill increases are now fixed?

A: Not necessarily. Large-load tariffs operate at the state public utility commission's retail rate layer. A significant share of data center-driven cost, at least in PJM's footprint, is showing up at the separate FERC wholesale capacity market layer — $13 billion in Q1 2026 alone, per PJM's own market monitor — which state tariffs do not directly address.

Q: If a data center pays 100% of its own infrastructure costs under a large-load tariff, can it still be raising my bill?

A: Yes, through a separate channel. New gas-fired generation built to serve data center load buys fuel on the same market as every existing gas plant. EIA's February 2026 Short-Term Energy Outlook modeled a high data-center-demand scenario and projected roughly a $0.50 per MMBtu increase in natural gas costs from that added demand. In wholesale markets like PJM's, generators are typically paid a single clearing price set by the last, most expensive unit dispatched to meet demand — the marginal unit — and gas plants routinely set that price. A higher gas price raises what every dispatched generator gets paid for that interval, not just gas plants, which is a cost no large-load tariff addresses, because tariffs govern infrastructure and demand charges, not how the wholesale market sets its clearing price.

Q: What's the difference between a state rate case and a PJM/FERC cost allocation? A: A state rate case, decided by a state PUC or PSC, governs distribution service and local generation costs within that state. FERC and regional grid operators like PJM govern the wholesale capacity and energy markets and transmission tariff layer, which operate independently of state retail rate decisions and can drive significant cost increases of their own.

Q: Why does the type of new generation — gas versus nuclear — matter for host communities, separate from cost allocation?

A: Employment. Historically, a 1 GW nuclear plant supports roughly 600 jobs at $100,000 to $120,000 annually, versus roughly 30 jobs for a comparable gas plant. Because gas turbines currently carry a five- to seven-year fabrication backlog, the faster-to-build generation to meet near-term demand is often not the option that leaves the host community with a comparable permanent job base. The data center itself is not a major source of local permanent jobs either — hyperscale facilities typically operate with as few as 20 to 30 permanent staff per 100 megawatts (so 200-300 staff for 1 GW load), and Virginia's Department of Taxation reported the industry added just 1,610 jobs statewide in fiscal year 2025 despite receiving $1.9 billion in state tax benefits over that period.

Sarah Gibboney, P.E. is the Founder of Gibboney Nuclear, PLLC, a nuclear licensing consultancy serving advanced reactor developers. She has 17 years of nuclear energy experience and has designed or licensed eight nuclear reactor designs.