How much are data centers costing electricity ratepayers in the PJM region?

NRDC and Synapse Energy Economics estimate that data centers drove $9.3 billion in capacity cost increases in a single year across PJM's 13-state territory. That translates to roughly $31 per month per household now, with projections reaching $70 per month by 2028 if current trends continue.

What is PJM's capacity auction shortfall and what caused it?

PJM's 2027-2028 Base Residual Auction produced the first-ever capacity shortfall: 6,625 MW below the reliability requirement. The grid procured 134,479 MW against a 141,101 MW target. Data centers account for 94% of projected demand growth, adding 5-7 GW annually while new generation supply adds only 2-3 GW.

What can ratepayers do about rising electricity costs from data centers?

Check whether your state has pending data center cost allocation legislation. Oregon passed the first dedicated data center rate class. Virginia's SB 253 would shift roughly $5.52 per month from households to data centers. Contact your state public utility commission and ask whether data center load growth is modeled separately in upcoming rate cases.

PJM's Electricity Prices Jumped 1,038% in Three Years. Data Centers Caused 63% of It. Ratepayers Got 100% of the Bill.

From $28.92 to $329.17. That is what happened to PJM Interconnection's capacity auction clearing price between the 2024-2025 and 2026-2027 delivery years. An increase of 1,038% in three years. Not gradually. Not predictably. PJM operates the largest electrical grid in North America, covering 65 million people across 13 states and the District of Columbia, and every one of those people pays the clearing price through their monthly electricity bill without having voted for it.

The 2027-2028 auction produced worse results still. Prices hit $333.44 per MW-day, slamming into the FERC-mandated price cap for the second consecutive year, and for the first time in PJM's history, the auction failed to procure enough capacity, leaving a 6,625-megawatt shortfall. PJM secured 134,479 MW against a reliability requirement of 141,101 MW, dropping the installed reserve margin to 14.8% against a 20% target.

NRDC and Synapse Energy Economics performed the decomposition, and the answer is unambiguous. Data centers caused 63% of the capacity price increase in the 2025-2026 auction alone, not as a contributing factor among many, but as the dominant driver of a crisis now affecting every ratepayer in 13 states.

The Per-Household AI Tax

Here is the calculation that matters, and the one that does not appear in PJM's investor presentations or in the press releases from data center operators announcing their next billion-dollar campus.

Synapse Energy Economics estimated $9.3 billion in data center-driven capacity costs for a single year across PJM territory, and with PJM serving approximately 25 million households, the arithmetic is straightforward. That is $372 per household per year, or $31 per month. Not a projection or a model, but an accounting of costs already allocated through the 2025-2026 capacity auction results.

Projections get worse from here, and considerably so. NRDC models show $70 per month per household by 2028 if the FERC price cap expires and demand growth continues at PJM's own forecasted trajectory. Cumulative costs through 2033 range from $100 billion (Synapse, conservative scenario holding the price cap) to $163 billion (NRDC, if the cap is removed). For comparison, PJM's annual capacity bill historically ran $8 to $10 billion, and it now runs $27 to $30 billion.

Delivery Year	Clearing Price ($/MW-day)	Change from 2024-25	Notes
2024-2025	$28.92	Baseline	Trough year
2025-2026	$269.92	+833%	Data centers caused 63% of increase
2026-2027	$329.17	+1,038%	Hit FERC price cap
2027-2028	$333.44	+1,053%	Hit cap again. First-ever shortfall: 6,625 MW

94% of Demand Growth, 0% of Infrastructure Cost

PJM's own long-term load forecast projects 32 GW of new peak demand between 2024 and 2030, and data centers account for 30 GW of that figure, which is ninety-four percent. New data center demand arrives at 5 to 7 GW per year. Supply adds only 2 to 3 GW per year. That deficit compounds annually with no mechanism to close it.

Under PJM's current cost allocation framework, transmission and capacity upgrades required to serve data center load growth are socialized across all ratepayers. A hyperscaler building a 500 MW campus in Loudoun County, Virginia does not pay for the grid upgrades that campus requires. Residential customers in Ohio, Maryland, and New Jersey do. They pay. They always pay. Seven mid-Atlantic states collectively absorbed $4.3 billion in data center connection costs charged to ratepayers in 2024 alone, according to filings reviewed by the PJM Independent Market Monitor.

Dominion Energy's zone in Virginia, the densest concentration of data centers on Earth, carries capacity fees 65% higher than the rest of the PJM grid. Maryland ratepayers saw increases of approximately $18 per month, Ohio ratepayers saw $16 per month, and none of these households operate data centers.

The Three Mile Island Paradox

While PJM scrambles for capacity, it is simultaneously blocking new supply from connecting. Constellation Energy's $1.6 billion project to restart Three Mile Island Unit 1, backed by a 20-year power purchase agreement with Microsoft, would add 835 MW of zero-carbon baseload generation to the grid. PJM told Constellation that transmission upgrades will not be ready until 2031.

Read that twice. PJM is 6,625 MW short, has proposed an emergency 15 GW auction because the situation is that desperate, and simultaneously told a willing generator with a signed billion-dollar contract and a functioning nuclear plant that 835 MW of carbon-free power cannot connect for five more years. That 835 MW represents 5.6% of what PJM is begging the market to provide through emergency procurement. Just 5.6%. A bureaucracy that created the crisis is now preventing its own solution. Nobody sees the irony.

The Emergency Auction

In April 2026, Trump and 13 bipartisan governors called for an emergency capacity auction targeting up to 15 GW of new generation, with an estimated price tag exceeding $15 billion. Its two phases work as follows: Phase 1 positions PJM as a matchmaker for bilateral contracts between data centers and generators, with terms of 2 to 15 years, while Phase 2 provides a central backstop procurement mechanism if Phase 1 falls short. A vote is expected in May 2026, with FERC submission in June and target online dates by June 2031, assuming permits and financing and construction timelines cooperate, which is an assumption that deserves heavy skepticism given how slowly and unpredictably American energy infrastructure typically moves.

Thirteen governors from both parties agreed on one thing: data centers must pay for the capacity they drive. That consensus is remarkable in 2026 American politics. Think about that. It also reveals how severe the problem has become when bipartisan agreement is the least surprising part of the story.

238 Bills in 50 States

State legislatures across the country are not waiting for FERC. According to Introl's legislative analysis, all 50 states introduced data center legislation in 2025, producing 238 bills total, of which 40 were enacted in 21 states.

Oregon's POWER Act (HB 3546) created the first dedicated data center rate class in the country, requiring 10-year minimum contracts and separate cost allocation so that residential customers do not subsidize commercial hyperscale loads. Virginia's SB 253 would shift distribution and capacity costs from residential ratepayers to data centers exceeding 25 MW. The Virginia State Corporation Commission estimated the savings at roughly $5.52 per month per household and a 15.8% rate increase for qualifying data centers, and Dominion Energy supports the bill.

Strongest Counterargument

Data centers bring massive economic benefits to the communities where they locate. Tax revenue, construction jobs, permanent operations positions, and the downstream economic activity generated by cloud computing services that millions of businesses and consumers depend on daily. AI services running inside those PJM-connected data centers are used by the same 65 million people whose electricity bills increased, and Microsoft Azure, AWS, and Google Cloud host the productivity tools, streaming services, and communication platforms that define modern work and life. Blocking or penalizing data center construction means falling behind in a global AI infrastructure race where China, the UAE, and Saudi Arabia are investing hundreds of billions.

This argument carries weight precisely because it is true, and data centers are not frivolous investments in a globally competitive technology landscape. Cost allocation, not the investment itself, is the problem. When a hyperscaler builds a campus that requires $500 million in grid upgrades, and a retired teacher in Trenton pays for those upgrades through her electricity bill, the economic benefits accrue to the corporation's shareholders while the costs accrue to the public. That is not an argument against data centers. It is an argument for making data centers pay their own infrastructure costs, which is exactly what 238 state bills and 13 governors are now demanding with increasing urgency.

Limitations

The $70 per month household projection depends on policy choices not yet made, particularly the FERC price cap expiration. If the cap stays, the trajectory is lower but still steep and still beyond anything in PJM's recorded history. NRDC and Synapse estimates diverge by $63 billion on cumulative costs, reflecting genuine uncertainty about demand trajectories and generation buildout rates. PJM's 94% data center share of demand growth is the grid operator's own projection, but actual construction depends on permitting, financing, and market conditions. A per-household calculation divides total capacity costs evenly across 25 million households, but actual bill impacts vary by state, utility zone, and rate structure. The PJM Independent Market Monitor confirmed the underlying cost dynamics at a Brookings Institution panel in March 2026, but granular data on which specific companies' data centers drive the most demand remains unavailable.

What You Can Do

If you live in PJM territory (Pennsylvania, New Jersey, Maryland, Virginia, Ohio, West Virginia, Delaware, Illinois, Indiana, Kentucky, Michigan, North Carolina, Tennessee, or DC): Your state likely has pending data center cost allocation legislation. Search your state legislature's bill tracker for "data center" or "capacity cost allocation." The bipartisan governor coalition's principles document provides specific policy language you can reference when contacting your state public utility commission.

If you work in energy policy or utility regulation: Oregon's POWER Act provides a legislative template for dedicated data center rate classes. The Virginia SCC's modeling methodology for SB 253 shows how to calculate the residential rate reduction from cost reallocation. Both are publicly available and adaptable to other state contexts.

If you work at or invest in companies building data centers: Track PJM's emergency auction proceedings closely. The May 2026 vote and subsequent FERC submission will determine whether bilateral contracts between generators and data centers become the standard procurement model, replacing the socialized cost structure. Companies that lock in bilateral PPAs now, as Microsoft did with Constellation, will have cost certainty. Companies relying on socialized grid capacity will face escalating political and regulatory pressure to pay their share.

If you care about climate implications: The TMI paradox reveals a genuine conflict. PJM needs 15 GW of new capacity, but the interconnection queue delays carbon-free nuclear by five years. The Department of Energy projects data centers will consume up to 12% of US electricity by 2028. Whether that power comes from renewables, nuclear, or natural gas depends entirely on how fast interconnection bottlenecks are resolved.

Bottom Line

PJM's capacity auction prices increased 1,038% in three years. Data centers caused 63% of the increase and account for 94% of projected demand growth. Sixty-five million people are paying for grid upgrades they did not request to serve data centers they do not operate. The per-household cost is $31 per month today and heading toward $70 per month by 2028. Meanwhile, 835 MW of zero-carbon nuclear power sits in an interconnection queue while PJM holds emergency auctions to find 15 GW. State legislatures in all 50 states are responding with 238 bills demanding that data centers pay their own way. The market created this problem. The question is whether regulators will fix the cost allocation before the cumulative tab reaches $163 billion, or whether 65 million ratepayers will continue subsidizing the AI boom on their kitchen tables.