Data centers are reshaping electricity costs most visibly in PJM, the grid serving 67 million people from Illinois to Virginia. PJM’s capacity price rose from $269.92/MW-day for 2025/26 to its $325–$333 cap for the next three delivery years, with annual auction costs around $16.4 billion; PJM’s independent market monitor attributes about $6.3 billion of 2028/29 capacity costs to data centers. Virginia’s JLARC estimated typical Dominion residential bills could rise $14–$37 a month by 2040 without rate changes. On the other side, studies find national retail prices broadly tracked inflation, data centers in Virginia currently pay their full cost of service, and new large-load tariffs increasingly make them pay for their own grid upgrades.
Do data centers raise electricity prices? The honest answer is: sometimes, in some places, through specific channels, and the outcome depends heavily on how regulators allocate costs. This article separates the evidence from the rhetoric, focusing on PJM’s capacity market, utility rate cases, large-load tariffs and the policy fight over who pays for new generation and transmission. It is part of our Data Center Economy hub; for the physical constraints behind the price signals, read why power is the new data center bottleneck, and for demand figures see our data center electricity consumption statistics.
Are data centers raising household power bills?
In tight markets such as PJM, rapid data center load growth has contributed to record capacity prices that flow into bills. Elsewhere, extra sales can spread fixed costs and hold rates down. Local market design decides which effect dominates.
How much of PJM’s capacity cost is linked to data centers?
PJM’s independent market monitor estimated that data centers account for about $6.3 billion of capacity costs for the 2028/29 delivery year, out of total auction costs of roughly $16.4 billion.
What are regulators doing about it?
Utilities and states are adopting large-load tariffs with minimum demand charges and long contracts, while FERC ordered grid operators in 2026 to prevent cost shifting from large loads.
Do data centers actually raise electricity prices?
They can, but not uniformly. Where large new loads arrive faster than new supply, as in PJM, they push up wholesale capacity and energy prices for everyone. Where utilities have spare capacity or charge data centers for upgrades, extra sales can spread fixed costs and hold rates steady.
Bills cover energy, capacity to meet peak demand, and network costs, and a large new customer affects each differently. It pays for the energy it uses and, under most tariffs, a share of network costs. But if its demand forces the system to procure scarce capacity or build new lines, the marginal cost of that expansion can exceed what it pays, and the difference is spread across all customers.
Evidence is mixed and contested. Lawrence Berkeley National Laboratory research on recent retail price trends found that national average retail prices largely tracked inflation, with some states rising faster and many seeing inflation-adjusted declines, and that drivers differ across states and over time. A causal study using Electric Power Research Institute data, as summarized in public reviews, associated higher data center capacity with slightly lower average residential rates from 2015 to 2024. The Data Center Coalition points to utilities in states such as Indiana and Georgia that it says may be lowering or freezing rates thanks to data center revenue.
The counter-evidence is concentrated in PJM. IIR News Intelligence, citing Heatmap News, reported that retail prices in PJM rose about 49% over five years against about 33% nationally, and consumer and environmental groups argue data center growth is the main reason PJM’s capacity market has cleared at its cap three years running. Both sets of facts can be true at once: national averages hide sharp regional differences.
Why have PJM capacity prices soared?
PJM capacity prices soared because forecast demand, driven largely by data centers, rose while retirements and slow interconnection limited new supply. The 2025/26 auction cleared at $269.92/MW-day, and the next three auctions hit or approached the FERC-approved price cap of about $325–$333.
PJM’s Base Residual Auction buys capacity commitments three years ahead. According to PJM’s 2027/28 auction report, the 2025/26 auction cleared at $269.92/MW-day at a total cost of about $14.7 billion; 2026/27 at $329.17/MW-day and $16.1 billion; and 2027/28 at the $333.44/MW-day cap, about $16.4 billion. By comparison, PJM said its 2024 auction, for 2024/25, cost about $2.2 billion. The 2027/28 report attributed a 5,249.9 MW increase in forecast load largely to additional large loads, and found that without the price collar the auction would have cleared at about $529.80/MW-day and cost $26.3 billion.
The 2028/29 auction, announced in July 2026, cleared at the $325/MW-day cap in every zone, procured about 138,317 MW and fell roughly 6.8 GW short of PJM’s reliability requirement, leaving a reserve margin around 14.7% against a target near 20%. PJM said uncapped prices would have reached nearly $555/MW-day and costs about $29.7 billion. Only around 525 MW of new or upgraded capacity cleared. Natural gas made up 46% of cleared resources, nuclear 20% and coal 18%, according to IIR.
The price collar, with a $175 floor and a cap around $325, was championed by Pennsylvania Governor Josh Shapiro and approved by FERC in 2025. In April 2026 FERC approved extending it to the 2028/29 and 2029/30 auctions. The cap limits consumer exposure, but critics argue it also mutes the price signal needed to attract new power plants.
| Delivery year | Clearing price ($/MW-day) | Approx. total cost | Note |
|---|---|---|---|
| 2024/25 | n/a | ~$2.2 billion | Pre-surge baseline (PJM) |
| 2025/26 | $269.92 | ~$14.7 billion | First big jump |
| 2026/27 | $329.17 | ~$16.1 billion | Collar in place |
| 2027/28 | $333.44 (cap) | ~$16.4 billion | Uncapped sim.: $529.80; reserve shortfall ~6.6 GW |
| 2028/29 | $325.00 (cap) | ~$16.4 billion | Uncapped sim.: ~$555; shortfall ~6.8 GW; DC share ~$6.3bn (IMM) |
Source: PJM 2027/28 Base Residual Auction report; PJM Inside Lines / Wikipedia for 2024 auction cost; Utility Dive, POWER Magazine and IIR News Intelligence (citing Monitoring Analytics) for 2028/29.
How much of the cost increase is due to data centers?
PJM’s independent market monitor, Monitoring Analytics, estimated that data centers planned for PJM account for about $6.3 billion of 2028/29 capacity payments. PJM itself attributes forecast load growth largely to large loads, primarily data centers, but declined to comment on the estimate.
Joe Bowring of Monitoring Analytics described the $6.3 billion as a cost imposed on all customers in the PJM footprint. The market monitor had already filed a complaint at FERC in November 2025 seeking to limit PJM’s ability to add large loads that cannot be served reliably. Separately, a 2026 Argonne National Laboratory review, as reported in public summaries, estimated about $4 billion in PJM high-voltage transmission costs tied to data center demand had been spread across all utility customers under existing planning rules.
PJM’s forecasts explain why the debate is so heated. An LBNL briefing for state consumer advocates noted that PJM’s 2026 forecast includes large-load adjustments of 39 GW by 2030 and 87 GW by 2040, against a system peak of around 154 GW. IIR reported that 1,321 planned data centers worth about $1.18 trillion sit in PJM’s footprint, roughly 41% of the planned US total.
The cost does not land evenly. A consultant at Unison Energy told Utility Dive that a 10 MW industrial customer’s monthly capacity charge would rise from about $6,000 in 2024 to roughly $70,000 in 2028. For households, the impact depends on how each state’s utilities pass capacity costs through, and on whether retail rates are set by regulators or by competitive suppliers.
Who pays for the transmission and grid upgrades data centers need?
Under traditional rules, many network upgrades are socialized across all customers, while directly attributable connection costs are charged to the new customer. Regulators are now shifting more costs onto data centers through dedicated rate classes, direct assignment and upfront payment requirements.
Virginia’s Joint Legislative Audit and Review Commission (JLARC), in its 2024 study, found that current Dominion rates allocate costs appropriately and that data center customers were paying their full cost of service. But it warned that rapid growth was still likely to raise costs for other customers through new fixed infrastructure, higher energy prices and more reliance on imported power. JLARC estimated a typical Dominion residential customer could see generation and transmission-related costs rise by about $14 to $37 a month by 2040, in constant dollars, and suggested a separate data center rate class among other options.
States are acting. LBNL’s 2026 Speed to Power briefing listed several cost allocation changes: Texas regulators replaced the 4CP transmission cost method with a 12CP approach in Docket 58484; Virginia’s State Corporation Commission ordered Dominion to explore alternative generation cost allocation; Oregon approved Portland General Electric’s peak growth modifier, which assigns costs partly by contribution to peak growth; Michigan required Consumers Energy to consider direct assignment to a new large-load class; East Kentucky Power Cooperative requires large loads to pay all costs they cause; and PG&E’s Electric Rule 30 lets customers that prepay transmission work connect faster.
What are large-load tariffs and do they protect other customers?
Large-load tariffs are rate classes for very large customers that typically require minimum monthly demand payments, contracts of 10 to 15 years, collateral and exit fees. They aim to ensure data centers pay for the infrastructure built for them even if their usage falls short.
LBNL’s briefing documented a spectrum of designs. Dominion Energy Virginia’s GS-5 class covers customers of at least 25 MW with high load factors; Florida Power & Light uses a 50 MW threshold; We Energies 150 MW; Arizona Public Service 5 MW; and El Paso Electric targets data centers above 50 MW. Thresholds range from no minimum to 500 MW, and contract terms are converging on 10 to 15 years. Oregon regulators approved a large-load tariff for PGE covering data centers above 20 MW in May 2026, and Ohio regulators approved a data center tariff for AEP Ohio in 2025 that requires new data center customers to pay for a minimum share of their contracted demand.
Not every tariff survives review. FERC rejected a large-load tariff proposed by Tri-State Generation and Transmission in October 2025. And in Delaware, McGuireWoods reported, the Public Service Commission paused new large-load interconnections while the utility develops a tariff, while state legislation would require a certificate to operate for loads of 30 MW or more.
Do these tariffs work? They reduce the risk of stranded network costs and discourage speculative requests, but they mostly address costs the utility incurs directly. They do less about market-wide effects, such as higher capacity prices in PJM, which arise because new demand tightens the whole system. That is why the debate has moved to wholesale market design and federal regulation.
What are FERC, PJM and states doing about cost shifting?
FERC’s June 2026 show-cause orders require six grid operators to adopt rules preventing cost shifting from large loads, including financial security and cost transparency. PJM is pursuing a reliability backstop procurement, and federal and state leaders have called for long-term contracts for new generation.
FERC’s orders of 18 June 2026, issued under Section 206 of the Federal Power Act, found that existing tariffs at PJM, MISO, SPP, CAISO, NYISO and ISO-NE appear unjust and unreasonable because they lack clear provisions for integrating large loads. One of five required reform areas is preventing cost shifting, including credit requirements and a pro forma cost-recovery agreement. Another is new transmission services for flexible loads willing to limit consumption during peaks, which could let data centers connect without forcing the grid to expand for their full demand.
Inside PJM, the 2028/29 shortfall prompted plans for a Reliability Backstop Procurement to close the roughly 6.8 GW gap, with an auction targeted for September 2026, and a separate “connect and manage” tariff for large loads that do not bring new generation. Generators including Constellation and Vistra proposed procuring the full requirement with 15-year contracts and a $420/MW-day cap. In January 2026, the Interior and Energy Secretaries and 13 governors issued a statement of principles supporting a backstop auction with 15-year generation contracts, which they said could support about $15 billion in new investment.
Environmental groups take a different view. The Natural Resources Defense Council argued after the July 2026 auction that PJM should remove large loads that have not brought their own new supply, make data centers pay for the plants and lines built to serve them, and speed interconnection of clean energy stuck in the queue. Data center companies respond that they are increasingly funding new supply directly, including nuclear restarts and uprates described in our guide to nuclear power for data centers.
Is the picture different outside PJM?
Yes. Texas uses an energy-only market without a capacity auction, so data center load shows up in energy prices and transmission costs rather than capacity charges. Vertically integrated utilities in the Southeast and West recover costs through regulated rates, where tariff design determines who pays.
In Texas, Senate Bill 6, signed in 2025, directed the Public Utility Commission to adopt rules on large-load interconnection, forecasting, emergency demand reductions, net metering with existing generation and transmission cost allocation, with several rulemakings running through 2026. ERCOT’s large-load queue has exploded to roughly 438 GW of requests, and its new batch study process is designed to prioritize projects with financing and land. Because Texas customers pay for transmission through regulated charges, the switch to 12CP cost allocation directly affects how much of that burden lands on large loads.
In vertically integrated states such as Georgia, utilities propose new generation in integrated resource plans and recover costs through regulated rates, so the size and design of data center rate classes determines the household impact. LBNL noted Georgia Power had more than 15 GW of large loads committed to receive service as of April 2026. Internationally, the debate looks different again: Ireland’s grid operator has restricted new data center connections around Dublin, while Gulf states market cheap power as an incentive, as covered in our article on the Gulf data center race.
What does this mean for businesses, investors and households?
For households and small firms, the key variable is local cost allocation, not national headlines. For data center developers, power costs and grid obligations are rising. For investors, regulated utilities and power generators benefit from load growth, while political risk is increasing.
Households in PJM states face sustained capacity costs through at least May 2029, cushioned by the price collar. Large commercial and industrial customers are more exposed, as the Unison Energy example shows. Developers should expect more minimum-take obligations, collateral and curtailment conditions, which raise the effective price of power and favour sites with spare capacity. Those cost pressures feed into the economics described in our article on data center operating costs.
For investors, the electricity price debate is a two-sided risk. Utilities with large rate-base growth and generators with existing capacity, especially nuclear and gas, benefit from tight markets. But political backlash, in the form of rate caps, tariff mandates or moratoria, can limit returns and slow data center development. That tension is one of the key questions in the wider AI data center bubble debate.
The most likely path is not a simple yes or no. Regulators are steadily moving toward rules that make data centers pay more of their incremental costs, bring their own supply or accept flexibility. If those rules work, data center growth could eventually help spread fixed grid costs. If they lag, ratepayers in the tightest markets will continue to absorb part of the AI build-out’s bill.
Frequently Asked Questions
Why do data centers affect my bill if I don’t use them?
Most power systems share the costs of capacity and transmission across all customers. When large new loads tighten supply, wholesale prices for capacity or energy can rise for everyone, and network upgrades may be partly socialized. In PJM, the market monitor estimated data centers account for about $6.3 billion of capacity costs for 2028/29, which are recovered from all load-serving customers.
Do data centers pay their fair share of grid costs?
It depends on the jurisdiction. Virginia’s JLARC found data centers were paying their full cost of service under Dominion’s rates, but warned future growth could still raise costs for others. Many utilities now use large-load tariffs with minimum demand charges and long contracts, and FERC’s 2026 orders require grid operators to address cost shifting, so the share paid by data centers is rising.
Will electricity prices fall if AI demand slows?
Not immediately. Capacity commitments are bought three years ahead, so PJM’s prices are largely locked in through May 2029. If data center demand later fell short of forecasts, capacity prices could ease, but customers could face stranded costs for infrastructure built for load that did not arrive. That is why regulators are requiring long contracts and exit fees from large loads.
Are data centers good for local economies despite power costs?
They can generate substantial property and equipment tax revenue and construction activity, although permanent jobs are modest relative to the capital invested. Whether the net local effect is positive depends on how energy, water and grid costs are allocated, which varies widely by state and utility, and on whether new tariffs make data centers pay for the infrastructure built to serve them.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.