It is the question I hear most often when a data center is proposed near homes. People are not really asking about megawatts. They are asking whether a building they did not invite is going to show up in the one bill they cannot skip. Their bills have gone up. A very large new customer has arrived in a lot of places at the same time. Putting those two facts together feels obvious.
The studies do not support the obvious version, at least not yet. They support a more specific one. Read in full, the evidence says data centers did not drive the rate increases households have already lived through, that in some places large steady load has actually held rates down, and that the forward risk in a few tight grids is real enough that regulators are now rewriting the rules before it lands on residential customers. Those are three different claims. Mixing them up is how this debate goes wrong.
Yes, your bill went up. That part is not in dispute
Start with what households actually experienced. U.S. residential electricity prices rose 29% from 2019 to 2025. Nobody needs a study to tell them that if they have opened the bill. The argument is only about attribution: how much of that increase belongs to data centers, and how much belongs to everything else a utility has been paying for in the same years, from fuel and financing to replacing old lines and hardening the grid against storms and fire.
That distinction matters for real estate. In my work on powered land, the price of a site now turns on deliverable power, and deliverable power turns on what the local grid can actually support. If new load were the main reason bills rose everywhere, the map of rate pain would follow the map of data center construction. It does not, cleanly. That is the first thing the studies show.
The correlation is real. The causation is not there
The National Association of Realtors looked at this at county level in its 2026 Data Center Impact Report. Residential electricity rates rose 21.4% from 2020 to 2024 in counties with 10 or more data centers, compared with 15.7% in counties without them. That gap is real, and opponents of new projects will quote it. They should also quote the rest of the finding. Counties with the most data centers do not have the highest residential rates, and the report does not find that the data centers caused the faster growth. Data centers tend to land in places that were already growing, already building, and already putting pressure on their grids for other reasons.
A study built to separate cause from coincidence points the other way. Researchers at the Electric Power Research Institute published causal work in June 2026 that tries to correct for the fact that data centers choose places with cheap power, which can hide their true effect in simple comparisons. Their estimate is that data centers caused average U.S. retail electricity rates to fall modestly from 2015 to 2024. The mechanism is not exotic. A grid carries large fixed costs whether it sells a lot of power or a little. A customer that draws power around the clock, at a high and steady load factor, spreads those fixed costs over far more kilowatt hours.
E3, the energy consulting firm, reached a similar conclusion from a different direction in its 2026 white paper on the drivers of rising rates. Reviewing 11 quantitative studies, E3 found no historical evidence of cost shifting from data centers to residential or small commercial customers under existing rate structures. In its own analysis of Amazon data centers across four utility territories, each site generated roughly $3.4 million in net surplus on average, meaning the revenue from the site ran ahead of the incremental cost of serving it. Readers should know two things about that study before they lean on it. It was funded by the Data Center Coalition, an industry group, though conducted by E3, and its authors describe history, not a guarantee about what happens next.
The clearest real world example of the surplus effect sits in California. PG&E has cut rates 11% since 2024, and its chief executive credits data center growth as part of the reason, because new large load brings in revenue against a system whose biggest costs do not rise in proportion. That is a utility executive making her own case, so weigh it accordingly. It still shows the direction the arithmetic can run when a system has room to absorb load.
Where the risk is real, and it is mostly ahead of us
None of that means a household in a tight grid is imagining things. The strongest evidence against data centers is not in the historical retail rate record. It is in the capacity market that prices tomorrow's power, in the largest grid in the country.
In PJM, the capacity auction for the 2025/2026 delivery year cleared at $269.92, up more than 9 times the prior auction level. Monitoring work on PJM auctions puts $6.5 billion, or 40%, of a $16.4 billion auction total as tied to data center load. Capacity is only one slice of a household bill, so an auction spike does not translate dollar for dollar into that bill. But when the price of simply having enough power available jumps that far, that fast, in a grid adding data center load faster than it is adding generation, some of that cost will reach ordinary customers unless regulators assign it elsewhere. That is not ideology. That is how cost recovery works.
There is also a documented leak in the current rules. The Union of Concerned Scientists has estimated that more than $4 billion in costs to connect data centers to the grid has fallen on regular customers through a connection cost loophole. A loophole is a different problem from load growth itself. It can be closed by tariff design, and increasingly it is being closed, which brings me to the part of this story that is moving fastest.
Regulators are now pricing the risk onto the data center
Watch what rate makers are doing, because they are no longer waiting for the argument to settle. In August 2026 the Tennessee Valley Authority took data centers out of its manufacturing rate class and put them in a dedicated rate of their own. The change leads to approximately 10% higher charges for data center power, phased in over three fiscal years for existing customers. New developments face upfront capacity commitment charges of about $1.5 million per MW, paid over three to five years. In plain terms, the data center now pays more of its own capacity cost up front, in the open, instead of that cost being averaged into everyone else's rate later.
Other states are using blunter tools. The Georgia Public Service Commission approved freezing Georgia Power base rates through 2028, a direct regulatory answer to load growth rate pressure. In Pennsylvania, the governor's deal with PJM was aimed at saving consumers more than $21 billion on utility bills. The tools differ. The direction does not. The cost of new capacity is being pushed toward the customers who cause it, earlier and more explicitly than it was five years ago.
For a landowner or a developer, that changes the underwriting, and it is why I include utility tariff structure in the checklist I use to underwrite a data center rather than treating power as a yes or no question. Two sites with the same megawatts on paper can carry very different costs once the tariff, the upfront capacity charge, and the timeline are priced in. That difference flows straight into what powered land is worth and how I value data center land, and it is one reason efficiency still matters: a building with a lower PUE draws less total power for the same compute, which is real money under a stricter tariff.
The bottom line
On the evidence available in 2026, data centers did not cause the residential rate increases of the last several years. Bills rose about 29% nationally from 2019 to 2025 for a stack of reasons, and the careful studies, including one designed to prove causation, find data center load holding rates flat to modestly down over the period they can measure, because steady around the clock load spreads fixed grid costs. The honest exception is forward looking and local. In PJM, capacity prices tied in part to data center load have spiked, a connection cost loophole has already shifted billions toward ordinary customers, and the next wave of load will arrive in grids with far less spare room than the grid that absorbed the last wave. The studies describe the past accurately. The new tariffs in places like the TVA service area tell you what regulators expect from the future, and they are making the data center pay first. That, for a homeowner, is roughly the right outcome, and for an investor, it is a cost line to underwrite now rather than discover later.
Sources: E3, Understanding the Drivers of Rising Electricity Rates and the Role of Data Centers (2026); Watten, Bistline, and Blanford, Have Data Centers Raised Your Electric Bill? Causal Evidence from the United States, Electric Power Research Institute (June 2026); National Association of Realtors, 2026 Data Center Impact Report; Union of Concerned Scientists, via E3 (2026); Monitoring Analytics, PJM capacity auction reporting, via the Data Center E-Primer (2026); Utility Dive, via E3 (2026); Tennessee Valley Authority rate action, via Data Center Dynamics (August 2026); Georgia Public Service Commission and Pennsylvania PJM agreement, via E3 (2026).