Regional Power Costs May Jump 57% in Server Hotspots

In concentrated hotspots, principally Virginia, the broader Mid-Atlantic corridor, the Ohio Valley, and West Texas, the increase could reach 57 percent

Decision Focus

A team from four universities, led by NC State’s Jeremiah Johnson, published a national hour-by-hour optimization model in Environmental Research Letters that maps what the data center build-out does to U.S. electricity costs and carbon emissions through 2030. The model runs 26 grid regions simultaneously and tests multiple fuel-cost scenarios. For operators managing energy procurement across U.S. markets, the finding concentrates in a specific place: regional cost exposure is already clustering in the same geographies where data center density is highest, and that overlap is about to intensify.

90-Second Brief

As the week closes, the model projects nationally averaged U.S. Electricity costs rising 6 to 29 percent by 2030 depending on scenario. In concentrated hotspots, principally Virginia, the broader Mid-Atlantic corridor, the Ohio Valley, and West Texas, the increase could reach 57 percent. Power sector carbon emissions could climb up to 28 percent relative to a scenario with no data center growth, reversing a meaningful share of reductions the U.S.

What Is Really Happening?

For nearly two decades, U.S. electricity demand was flat enough that utilities could plan on stable load growth. Large-scale computing loads — data centers and cryptocurrency mining — broke that pattern. This study is notable for quantifying the full downstream effect on both prices and carbon emissions simultaneously at hour-by-hour resolution, which prior national analyses had not done.

The key structural dynamic is transmission concentration. Data centers do not spread load evenly; they cluster near existing fiber, power infrastructure, and favorable land-use regimes. Virginia’s dominance in the model reflects this directly: the existing density of server farms in Northern Virginia is high enough that the model shows coal plants in that region running harder to meet incremental server demand — a directional outcome few operators’ sustainability teams modeled five years ago.

The gas price paradox the study surfaces is counterintuitive and operationally relevant. In most energy scenarios, cheap natural gas decarbonizes the grid by displacing coal. With high, constant data center loads added, cheap gas instead feeds server demand, leaving coal plants running at baseline. Expensive gas produces the opposite result: renewables fill more of the gap and emissions fall. Operators whose long-term PPAs carry embedded gas price assumptions may face outcomes that diverge from sustainability targets depending on which price path materializes.

Why It Matters for Global Heads of Data Center Energy

The 57 percent figure is a worst-case regional scenario, not a national average, but it identifies precisely where tariff exposure will concentrate. For portfolios with significant Virginia or Mid-Atlantic footprint, this study provides quantified signal that utility rate trajectories in those markets are moving structurally against you — not a cyclical fluctuation hedgeable through a VPPA, but a load-growth dynamic that utilities will eventually socialize across the rate base or redirect onto large industrial customers through new tariff structures.

The carbon reversal finding creates a Scope 2 exposure that sits outside most operators’ current sustainability models. Two decades of U.S. grid decarbonization that operators assumed as a background tailwind for market-based REC strategies may be partially offset by the same load growth their own facilities are contributing to. This does not invalidate REC or VPPA strategies on their own terms, but it significantly strengthens the operational case for additionality requirements and 24/7 CFE matching over annual average accounting — approaches that demonstrate new clean capacity was actually built rather than relying on existing clean generation being recredited into a deteriorating grid mix.

The distributional finding also carries practical weight for near-term site selection. Spreading new data centers across more states softens the worst regional price spikes without materially changing national averages. That is not an argument against Virginia or Texas in isolation, but it is a quantified basis for weighting power-cost trajectory stability more heavily in site-selection criteria, particularly as transmission congestion in hotspot regions compounds.

Forward View

Three fronts warrant active tracking over the next 18 months. First, whether federal renewable incentives — which the study identifies as capable of tempering both cost and emissions outcomes by redirecting new capacity toward wind and solar — are restored, modified, or eliminated in the current legislative cycle. The difference between incentive and no-incentive scenarios is large enough to affect forward PPA pricing assumptions in key markets.

Second, whether Virginia and PJM regulators move to socialize data center load-growth costs across the broader ratepayer base or begin requiring large industrial loads to absorb more of their interconnection and grid reinforcement costs directly. Either path changes the economics of hotspot-heavy portfolios, but in different directions and on different timelines.

Third, the gas price trajectory through 2027. Given the counterintuitive emissions-to-gas-price relationship the model identifies, market movements your trading desk tracks for cost management carry a non-obvious carbon directional implication that sustainability reporting will need to reconcile.

What Is Still Uncertain

The study is a modeling exercise, not a forecast. The 57 percent figure represents one scenario endpoint, not a central estimate. The model does not capture demand response commitments, behind-the-meter storage deployment at scale, or negotiated large-load tariffs that operators may access in specific markets. It also does not resolve how quickly interconnection backlogs clear or new transmission is built — both of which materially affect regional hotspot dynamics in either direction. The cryptocurrency mining load assumption carries a wide uncertainty band, cited in the study as between one-half and two percent of total U.S. electricity use as of the study period, a range broad enough to shift regional projections meaningfully. Finally, the emissions-to-gas-price relationship is scenario-dependent; the study tests several futures rather than converging on one.

One Question for Your Team

Does your current Scope 2 accounting and PPA portfolio assume continued U.S. grid decarbonization as a background condition — and if so, which commitments carry the most exposure if that assumption reverses in your primary markets before 2030?