July 8, 2026 • By Priya Greenwood

Your Builder Modeled Your Energy Bill at 16 Cents a Kilowatt-Hour. Ohio Just Hit 19.5.

Somewhere in a suburban Columbus sales office, a laminated sheet still sits on the counter showing prospective buyers that their new HERS 55 home will cost roughly $180 a month to heat, cool, and power. The number comes from an energy model that a RESNET-certified rater ran through approved software, fed with local weather data and the home's insulation specs and HVAC efficiency ratings and window U-factors and everything else the building science community has spent decades learning to measure with admirable precision.

One thing the model did not measure with precision was what electricity would cost when the family moved in.

Between April 2025 and April 2026, Ohio's average residential electricity price jumped from 16.32 cents per kilowatt-hour to 19.49 cents, according to the U.S. Energy Information Administration's Electric Power Monthly—a 19.4 percent increase in twelve months, not over a decade, not adjusted for inflation, just the raw distance between one April and the next, a span so short that the family buying the home in that Columbus suburb may not have even closed on their mortgage before the rate assumption underlying their energy model became a historical artifact.

The laminated sheet on the counter does not know this, and nothing in the sales process will update it.

The Grid Beneath the Model

What happened in Ohio happened across the thirteen-state footprint managed by PJM Interconnection, the largest regional transmission organization in the United States, serving 67 million people from New Jersey to Illinois. Pennsylvania's residential rate climbed 13.2 percent. New Jersey's, 16.8 percent. The pressure radiated outward into neighboring grids: New York, managed by NYISO, rose 14.6 percent. New Hampshire, under ISO-New England, saw a 15.1 percent spike that pushed its average residential price to 27.24 cents per kilowatt-hour.

The driver is not a weather anomaly or a pipeline disruption or a refinery fire. It is data centers, consuming electricity at a scale that reshapes the economics of everything else connected to the same wires.

PJM's capacity auction, the forward market where generators are paid to guarantee they can produce electricity during peak demand, cleared at $28.92 per megawatt-day for the 2024–2025 delivery year. For 2025–2026, that price exploded to $269.92 per megawatt-day. For 2026–2027, it hit the FERC-approved cap of $329.17 per megawatt-day, a ceiling that exists only because regulators feared what would happen without one. In the BGE and Dominion zones, northern Virginia, where the world's largest cluster of data centers draws power, prices had already reached $466 per megawatt-day the year before.

PJM itself projects 32 gigawatts of new demand by 2030. Thirty of those gigawatts, 94 percent, will come from data centers. Not residential load. Not manufacturing. Data centers. When analysts at ICF told Reuters on July 3 that PJM has no spare capacity to support new demand beyond next year, the statement landed not as warning but as arithmetic.

Three days later, PJM ordered emergency demand curbs as consumption approached 166.3 gigawatts, threatening to break a record set twenty years ago, and wholesale prices in northern Virginia spiked to $2,500 per megawatt-hour against a backdrop where $40 is considered normal.

What a HERS Rating Actually Promises

A HERS Index score is an elegant instrument for what it actually measures. A home rated HERS 55 uses roughly 45 percent less energy than the RESNET Reference Home, a standardized benchmark calibrated to the 2006 IECC. The measurement is denominator-agnostic: it tells you how much energy the home will consume, in kilowatt-hours and therms, given local climate and the home's physical characteristics. It says nothing reliable about what those kilowatt-hours will cost.

This is not a flaw in the system but a design choice that made sense when electricity rates were boring, moving a percent or two per year in either direction, predictable enough that any reasonable rate assumption would land in the neighborhood of reality over a mortgage's life.

That assumption just evaporated across PJM's entire footprint, and the evaporation was not gradual.

D.R. Horton, the nation's largest homebuilder, includes a disclosure on its HERS materials that most buyers probably skim past: "D.R. Horton does not warrant or guarantee the monthly or annual utility bills for the home." The fine print lists the reasons energy costs might diverge from the model: occupant behavior, weather, maintenance, appliance usage, indoor temperature preferences. Buried in that list, almost as an afterthought: "utility rates and fees."

That afterthought is now the dominant variable in whether a buyer's energy budget survives contact with reality.

Running the Numbers Nobody Ran

Consider a new all-electric home in the Columbus, Ohio, metro area. Eighteen hundred square feet, heat pump HVAC, heat pump water heater, induction cooktop, no gas line. HERS 55. The builder's energy model, calibrated with 2024 rate data, estimated annual electricity consumption at roughly 15,000 kilowatt-hours and annual cost at approximately $2,250, using Ohio's then-prevailing residential rate of around 15 cents per kilowatt-hour.

At April 2026's actual rate of 19.49 cents, that same 15,000 kilowatt-hours costs $2,924. The delta is $674 per year—30 percent above the modeled figure. Over the first five years of a thirty-year mortgage, assuming rates hold at 19.49 cents (PJM's auction trajectory suggests they won't fall), the cumulative unmodeled cost reaches $3,370. If rates climb just 3 percent annually from here, roughly half the historical pace for PJM states over the past two years, the total gap between what the buyer was told and what they actually pay exceeds $30,000 over the mortgage's life.

New York is worse. A comparable home consuming 12,000 kilowatt-hours at the state's April 2026 residential rate of 29.45 cents per kilowatt-hour would run $3,534 per year. If the builder modeled it at 2024's rate of approximately 24 cents, the buyer was told to expect $2,880. Annual surprise: $654. Over thirty years: you can do the arithmetic, and it is not comfortable.

For a buyer in Pennsylvania, the gap is $375 per year. New Jersey, $507. These are not theoretical projections fed through Monte Carlo simulations. They are the difference between EIA-published rates from April 2025 and April 2026, applied to the same kilowatt-hours.

The Counterargument, Stated Honestly

Three objections deserve full engagement.

First: efficient homes still save more than inefficient ones regardless of rate levels, and this is true and important. A HERS 55 home using 15,000 kilowatt-hours at 19.5 cents still pays less than a code-minimum home using 22,000 kilowatt-hours at the same price, and higher rates actually improve the return on efficiency investments because every kilowatt-hour you don't consume is worth more when electricity is expensive. Efficiency is not the problem; the gap between modeled and actual costs is.

Second: energy models predict consumption, not cost, and rate volatility is the buyer's responsibility. This is technically correct and practically useless, because no builder markets a HERS rating as "your home will consume 15,000 kilowatt-hours." They market it as "your estimated monthly energy cost is $180." The moment consumption gets translated into dollars on a sales sheet, the rate assumption becomes a promise, and the promise is expiring faster than the limited warranty.

Third: some states are successfully isolating data center costs from residential ratepayers, and Oregon's Public Utility Commission hiked data center rates 29 percent on July 7 while cutting residential rates by 1.3 percent, demonstrating that rate-class separation is achievable when regulators have the political will and the statutory authority to act. Tennessee passed legislation requiring data centers consuming 50 megawatts or more to pay for their own electrical infrastructure. If every state in PJM territory followed this model, the residential rate shock would attenuate considerably, but PJM's governance structure spans thirteen states and the District of Columbia, and coordinating policy across that footprint moves roughly as fast as building a new transmission line, which is to say: years, during which the bills keep arriving monthly.

What AI Could Fix, and What It Cannot

There is an obvious intervention point here, and it involves software, not concrete.

RESNET's approved modeling software (EnergyGauge, REM/Rate, Ekotrope) could integrate real-time or near-real-time utility rate data sourced from EIA's monthly publications rather than relying on static assumptions baked in during the rating. Cove.tool already does something adjacent in commercial energy modeling, pulling utility rate structures dynamically. Adapting this for residential HERS ratings would require RESNET to update ANSI/RESNET/ICC Standard 301, which defines how energy costs are calculated within the HERS framework. That standard was last updated with Addendum D-2022, adding a carbon index. Rate responsiveness was not on the agenda.

More ambitiously, AI-driven energy cost forecasting could replace static rate assumptions entirely. Machine learning models trained on EIA rate history, PJM auction data, regional generation mix, and data center pipeline announcements could produce probability-weighted cost ranges rather than single-point estimates. Instead of "$180 per month," a buyer would see "$165 to $240 per month through 2031, with a 30 percent probability of exceeding $220 in PJM territory." That is less pleasant to print on a laminated sheet. It is considerably more honest.

What software cannot fix is the political economy. Twenty years of grid underinvestment, a data center buildout that moves faster than transmission permitting, and a regulatory framework designed for a world where demand grew at 1 percent per year rather than the 4 to 5 percent EIA now projects. None of this yields to a better algorithm.

The Disclosure Gap

The EIA released its Short-Term Energy Outlook today, July 8, projecting that U.S. power consumption will climb from a record 4,195 billion kilowatt-hours in 2025 to 4,269 billion in 2026 and 4,399 billion in 2027. Commercial-sector demand, which includes data centers, will outpace residential demand this year for the first time in the EIA's history of tracking the data. The structural pressure on rates is not a blip. It is a trajectory.

Yet nothing in current residential disclosure practices requires a builder, rater, or real estate agent to flag that the energy cost estimate on a new home's marketing materials was calibrated to a rate environment that may be obsolete by the time the buyer files their first utility payment. Energy code compliance under the IECC doesn't address rates at all; it measures building performance in energy units, as it should, and leaves cost translation to the marketplace, where incentives run entirely toward optimism.

A home that genuinely uses 45 percent less energy than code minimum is a genuine achievement. A buyer who moves into that home expecting $180 a month and discovers $235 has not been defrauded. No law was broken, no standard was violated, no rater made an error. They were simply given a number generated by a system that assumes the grid beneath the home will behave the way it has always behaved, delivered at the precise historical moment when it stopped.

What You Can Do About It

If you are buying a new home in PJM territory or any region with significant data center construction, ask the builder or rater one question before signing: what electricity rate did the energy model use, and when was it last updated? Then check the EIA's Electric Power Monthly for your state's current average residential rate. If the model used a rate more than 10 percent below today's published figure, the estimated monthly cost on that sales sheet is already stale. Ask for a new one.

Builders: provide rate-sensitivity ranges alongside point estimates. Show the buyer what their annual energy cost looks like at the modeled rate, at the current EIA rate, and at a rate 20 percent above current. It costs nothing to produce. It inoculates you against the uncomfortable phone call eighteen months after closing.

For buyers who want structural protection against rate volatility, rooftop solar with battery storage remains the most effective hedge. It is not cheap. But in PJM territory, where capacity auction prices suggest sustained upward pressure on rates, the payback period on a $20,000-plus solar-and-storage installation is compressing faster than anyone modeled two years ago.

Limitations of This Analysis

This cost comparison uses EIA's April 2026 state-average rates, which are the most recent published monthly data. Summer rates in PJM territory may be significantly higher once July and August data are reported, given the record demand and price spikes observed in the first week of July. The "modeled cost" figures are estimates based on typical rate assumptions used in energy modeling software during 2024; actual modeled costs vary by rater, software version, and the specific utility rate schedule entered. Rate increases are most acute in PJM territory; states outside this footprint, particularly those with vertically integrated utilities and smaller data center exposure, may see more modest changes. This analysis does not account for on-site solar generation or battery storage, either of which can substantially insulate a homeowner from rate volatility, at an additional installed cost of $18,000 to $30,000 that the energy model also did not include.