A 23 billion dollar cost shift
Data center growth has pushed electricity prices up by about 23 billion dollars across the PJM market, according to an analysis published July 14 by Theodore Kury, director of energy studies at the University of Florida, writing for The Conversation and republished by Fortune. PJM is the grid operator covering all or part of 14 mid-Atlantic and Midwest states, and Kury argues the price increases already baked into its capacity auctions will persist through at least the end of 2028. The figure quantifies a tension that CTOs building or leasing data centers increasingly cannot ignore: the power their facilities draw is reshaping electricity markets and drawing political scrutiny.
The mechanism runs through PJM's capacity market, where the grid pays generators to guarantee supply during peak demand. Surging data center load has tightened the balance between supply and demand, and the most recent auctions cleared at sharply higher prices as a result. Those costs pass through to every customer connected to the grid, including households that never signed a data center lease. Kury's analysis frames this as a wealth transfer, with residential and small-business ratepayers absorbing charges triggered largely by hyperscale compute. For infrastructure leaders, the number is a warning that cheap grid power is becoming a contested political resource across the region.
How costs get shifted
A central point in the analysis concerns coincident peak demand, the method utilities use to allocate grid costs based on consumption during system-wide peaks. Kury notes that data centers can often fine-tune their electricity use minute to minute, dialing consumption down during the specific peak windows that determine their cost share in ways residential users cannot replicate. By shaving load precisely when the meter that matters is running, a sophisticated operator can reduce its allocated share of infrastructure costs, shifting more of the burden onto customers with less flexible demand. The result is a system where the largest users may pay less than their proportional impact on the grid.
This dynamic has direct implications for how data center operators are perceived by regulators and the public. As the cost shifting becomes visible, states are responding with moratoriums, special tariffs, and demands that large loads fund their own grid upgrades. New York paused new hyperscale projects above 50 megawatts, and other jurisdictions are drafting rules that require data centers to pay for the transmission they trigger. For CTOs, the era of assuming abundant, cheap, politically uncontested power is closing. Power strategy is migrating from a facilities footnote to a board-level risk that affects siting, cost forecasts, and community relations.
Where the pressure is worst
The analysis points to states including Texas, Pennsylvania, Nevada, Georgia, and New York as flashpoints where data center load is straining local grids and lifting bills. These are the same markets that attracted hyperscale investment for their land, incentives, and historically cheap power, and the influx is now testing the assumptions that drew developers in. Pennsylvania sits inside PJM and feels the capacity price increases directly, while Texas runs its own grid under ERCOT with separate but related stress. The geographic spread shows this is a structural feature of concentrated AI buildout rather than a quirk of any single utility's rate design.
For enterprise buyers evaluating providers, grid exposure is becoming a differentiator worth probing. A provider concentrated in a strained PJM zone faces more regulatory and price risk than one drawing on dedicated generation or sited in a region with slack capacity. That risk can surface as higher pass-through energy charges, delayed capacity due to interconnection queues, or reputational drag from local opposition. Asking a hyperscaler or colocation provider where its power comes from, and whether it funds its own grid upgrades, is a reasonable diligence question. The answer increasingly shapes both the cost and the reliability of the compute a buyer is contracting for.
Behind-the-meter as the escape hatch
The cost-shift problem partly explains why hyperscalers are racing to build dedicated, behind-the-meter generation. When a data center draws power from its own on-site plant rather than the public grid, it sidesteps the capacity market charges that Kury describes and insulates itself from rising retail rates. That logic is driving the wave of gas turbine orders, nuclear power purchase agreements, and co-located generation deals announced through 2026. It also lets operators claim they are adding load without burdening existing ratepayers, a politically valuable argument as public anger over bills grows across PJM and beyond into neighboring markets.
The behind-the-meter shift carries its own tradeoffs that buyers should understand. On-site gas generation raises emissions and complicates sustainability reporting, while dedicated nuclear or renewable supply takes years to deliver. Projects that bypass the grid can still draw scrutiny over water use, air permits, and local land impact. For CIOs with carbon commitments, a provider's power source now affects reported Scope 2 and Scope 3 footprints in material ways. The cleanest read of Kury's analysis is that power sourcing has become a first-order attribute of any data center decision, influencing cost, carbon, regulatory exposure, and the durability of the underlying supply.
What leaders should do
The practical response to this analysis is to treat power as a core variable in every infrastructure decision rather than a background utility. Technology leaders sourcing capacity should ask providers to disclose their grid exposure, their generation mix, and how they fund the transmission upgrades their load requires. Contracts should account for the possibility of rising pass-through energy costs and new large-load tariffs, both of which are spreading across regulated markets. Forecasts that assumed stable electricity prices need revisiting, because the PJM capacity increases show how quickly the energy line item can move against a multi-year data center commitment.
There is a reputational dimension leaders cannot delegate. As reporting like Kury's makes the ratepayer impact of AI concrete, enterprises associated with data center expansion face questions about whether they are shifting costs onto communities. Getting ahead of that scrutiny, by favoring providers that fund their own grid impact and disclose their power sourcing, protects both budget and brand. The 23 billion dollar figure is a preview of a debate that will intensify as AI load keeps growing. Building power strategy and community impact into procurement now is cheaper than retrofitting it after a moratorium or a rate backlash lands.
