AI, data centers, and your bill
ISO capacity is one knob among many—and its retail bite is dependent on location, product, and timing, not a blanket surcharge on everyone.
When news articles spotlight wholesale prices rising near big data-center clusters, that is a very real phenomenon in a few hubs - primarily in beleagured FERC capacity market areas. Whether retail bills rise—and by how much—depends on state rate design, cost allocation, how flexibly those large loads operate, and the engineering of the grid. [1]
Capacity markets are a separate retail pressure point in markets like PJM: recent capacity auctions cleared high, which raises one component of retail costs for LSEs serving that footprint. Capacity markets are a separate, regional pressure—only where they exist and only to the extent you’re exposed. In PJM, the 2026/27 Base Residual Auction cleared at the $329.17/MW-day cap (up ~22% from 2025/26’s $269.92), after some LDAs like BGE and Dominion cleared even higher the prior year. That can raise one component of retail costs for LSEs that pass RPM through, and PJM itself estimates on the order of ~1.5–5% bill impact for some customers. But that’s not universal: exposure varies by footprint (PJM vs. non-PJM), by product choice (RPM vs. FRR/hedged supply), by timing of default-service procurements, and by state allocation rules. In other words, capacity is one knob among many—and its retail bite is location-, product-, and timing-dependent, not a blanket surcharge on everyone.
AI, data centers, and your bill
Bloomberg spotlights wholesale prices rising near big data-center clusters. That’s real in a few hubs. Whether retail bills rise—and by how much—depends on state rate design, cost allocation, and how flexibly those large loads operate. If you need to understand why your bill is going up, you need to open up your bill and match it to the publicly available tariff sheets that your electricity bill “codes” to.
Bloomberg’s article on data center/AI costs and power bills, maps hubs where wholesale prices have risen sharply and ties that to the AI/data-center build-out.[2] That’s a wholesale story. It isn’t the same as a uniform retail outcome for households statewide. Nationally, average retail prices have risen only modestly in the last year.[3]
Wholesale vs. retail
Here’s the pass-through chain:
Locational wholesale costs (energy, congestion, in some markets capacity) move with local constraints and peak hours.[2]
Utility portfolios and riders (fuel, purchased power) translate some of that movement into revenue requirements, subject to timing and procurement choices.[4]
Rate design allocates costs across classes. Where regulators create a data-center class with tailored charges and commitments, general customers need not absorb the same share of grid upgrades or peak costs.[5][6] Look for rate class allocators to determine who pays what. Where regulators create a data-center class with tailored charges and commitments, general customers don’t have to carry the same share of upgrade or peak costs. (Virginia is moving this direction now).
Recent bill changes typically blend several forces: fuel/power costs, wildfire and storm hardening, interest rates and deferred capital, plus new large loads in specific places and hours.[3][4] Data centers are an additional and regional driver, not a universal one.
Where policy is adjusting
Virginia is the live case: a proposed data-center rate class and related measures to align cost causation with cost allocation for very large loads (e.g., threshold MW levels, term commitments).[5][6] Other jurisdictions are considering similar steps in ongoing planning and rate proceedings.[3]
The missing lever: dispatachability around peak demand windows
Flat, always-on megawatts are expensive at the margin. Curtailable, nodally dispatchable load paired with storage can reduce congestion rents, defer upgrades, and earn revenue by showing up in the right hours—not all hours. Treating the problem as temporal (a few peak hours) and topological (a few constrained interfaces) is certaintly a generation capacity cost solution: cheaper than broad socialization of costs and better for everyone’s bill. On the other hand, building firm upgrades that help electrify everyone’s homes and businesses with data centers or other large industrial loads subsidizing the buildout of that transmission backbone - is shared and socialized infrastructure. Bringing on transmission upgrades with curtailment solutions enabled so that loads do not contribute to phantom upgrade forecast, is a key solution as well.
What to watch for in your bill/utility policy
Rate-class outcomes. Do class-specific allocators and commitments actually shield residential/commercial customers?[5][6]
Targeted investments. Are upgrades and non-wires/storage solutions scoped to the actual constraints instead of spread system-wide?
Operating behavior. Do large loads enroll in curtailable/dispatchable programs and document measurable relief in constrained hours? Or, are they subject to new tariff structures that ring-fence the additive cost causation they bring to the system due to their firm capacity request?
If you want a hair-tearing evening, call your parent in another state and try to read their utility bill! Bills are rising because rate formulas allow them to rise. State commissions don’t “peg” rate increases to inflation or cost-of-living. They approve a utility’s revenue requirement in rate cases (rate base × allowed return + O&M + depreciation + taxes). That sets base rates—and it’s cost-of-service based on what the utility spent, not CPI-of-service based on what specific customers can actually afford in purchase power terms.
Endnotes
[1] See generally Bloomberg, discussion of wholesale price movements near data-center clusters and pass-through claims. AI Data Centers Are Sending Power Bills Soaring (Graphics/Big Take, 2025), https://www.bloomberg.com/ (last visited Sept. 29, 2025).
[2] Id. (mapping hubs with “as much as 267%” wholesale increases vs. five years prior and attributing part of the rise to data-center growth).
[3] U.S. Energy Information Administration, Electric Power Monthly & Electricity Monthly Update (2025), https://www.eia.gov/electricity/ (national average end-use price trends and state/sector tables) (last visited Sept. 29, 2025).
[4] U.S. Energy Information Administration, Electric Retail Price and Revenue Data; Rate Case and Fuel/Rider Materials by State (compilation landing pages), https://www.eia.gov/ (illustrating how fuel riders and purchased-power costs flow to retail) (last visited Sept. 29, 2025).
[5] Dominion Energy, Press Release, Dominion Energy Proposes New Rate Class for High-Load Customers (Apr. 1, 2025), https://www.dominionenergy.com/ (summarizing proposed data-center class) (last visited Sept. 29, 2025).
[6] Virginia Mercury, Dominion proposes higher utility rates, new rate class for data centers (Sept. 3, 2025), https://www.virginiamercury.com/ (reporting on the filing and stakeholder positions) (last visited Sept. 29, 2025).