Electricity demand in the United States is rising after more than a decade of relatively flat growth, with data center expansion emerging as one of the fastest-growing sources of new load in several regional power systems.
Large technology companies including Microsoft, Amazon Web Services, and Google continue to expand hyperscale data center capacity across major U.S. hubs such as Northern Virginia, Texas, and Ohio, according to utility planning documents and corporate disclosures. Northern Virginia remains the largest concentration of hyperscale data center infrastructure globally.
Utilities are increasingly treating this shift as a structural change rather than a marginal increase in demand.
Dominion Energy, in filings with Virginia regulators, has identified large-load customers—primarily data centers—as a key driver of long-term demand growth projections. Similar patterns have been noted by other utilities serving rapidly growing digital infrastructure regions.
According to the U.S. Energy Information Administration (EIA) 2025 Annual Energy Outlook, electricity demand is expected to grow roughly 1–2% annually through the late 2020s across core scenarios, reversing a long period of near-flat consumption. The EIA notes that data centers are becoming a meaningful contributor to incremental demand, although regional outcomes vary significantly depending on transmission capacity, generation mix, and local policy.
Demand growth is increasingly colliding with grid constraints
Regional grid operators have also highlighted rising pressure in their planning processes.
PJM Interconnection, which operates the largest wholesale electricity market in the U.S., has reported a growing backlog of interconnection requests for both generation and large-load customers. These queues now include renewable projects, storage systems, and hyperscale data center connections competing for limited transmission capacity in constrained zones.
In PJM’s most recent capacity auction for the 2025/2026 delivery year, clearing prices increased significantly compared to prior years. PJM attributes this to tightening reserve margins, changes in supply availability, and updated demand forecasts. Some market participants and analysts also point to growing expectations of large-load growth, including data centers, as part of forward-looking uncertainty.
Why “clean energy becoming more expensive” is mostly a system-level effect
A key distinction emerging in grid economics is the difference between:
- Generation cost (cost of producing electricity)
- Delivered system cost (cost of delivering electricity to where it is needed)
While solar and wind generation costs continue to decline, the delivered cost of clean electricity is increasingly shaped by:
- interconnection queue delays
- transmission expansion timelines
- congestion in high-demand regions
- required grid upgrade investments
In multiple FERC proceedings, long interconnection delays have been identified as a structural bottleneck in connecting new generation to load centers.
As a result, even when clean energy is cheap to generate, getting it delivered to fast-growing demand regions can involve significant additional system costs and delays.
The timing mismatch between data centers and grid expansion
This issue is becoming more visible as AI infrastructure scales.
Hyperscale data centers are typically built within 18–24 months. By contrast, major transmission projects can require 5–10 years due to permitting, land use, regulatory review, and multi-jurisdiction coordination.
This timing mismatch is now influencing both utility planning and corporate procurement strategies.
Microsoft, Amazon, and Google have each expanded long-term renewable procurement through power purchase agreements (PPAs), as reflected in their sustainability reports and regulatory filings. However, these same disclosures also note that grid congestion and interconnection constraints can affect the timing and location of actual clean energy delivery relative to contracted volumes.
Utilities are beginning to adapt pricing and planning models
Utilities including American Electric Power (AEP) and Dominion Energy have indicated in regulatory filings that unusually large individual customer requests—primarily from data centers—are now influencing transmission and distribution planning more directly.
In response, some utilities are evaluating revised tariff structures or “large load” categories to better allocate infrastructure upgrade costs associated with hyperscale demand.
This reflects a broader shift in grid planning: large new loads are increasingly treated as core drivers of infrastructure expansion, not marginal additions.
Not all forecasts agree on magnitude
There remains significant uncertainty around the scale of future electricity demand growth.
Some analysts argue that data center demand is highly concentrated and may be partially offset by:
- improvements in chip efficiency
- workload optimization
- cooling system advances
Others suggest that AI adoption trajectories could accelerate demand beyond current utility baseline assumptions, particularly as inference workloads scale globally.
This uncertainty is reflected in the wide range of demand scenarios used by utilities and regulators.
International comparison: coordinated expansion vs fragmented constraints
China continues to expand both renewable generation and ultra-high-voltage transmission infrastructure at large scale.
According to the International Energy Agency (IEA), China remains the global leader in annual additions of both generation capacity and transmission infrastructure. Its more centralized planning approach allows generation and grid expansion to be more tightly coordinated in many regions.
While system structures differ significantly, the IEA notes that this coordination can reduce some of the bottlenecks seen in more fragmented electricity markets.
Bottom line
The interaction between AI-driven electricity demand and U.S. grid constraints is increasingly shaping how new energy systems function in practice.
The key issue is not simply whether enough electricity is being generated, but whether it can be:
- connected
- transmitted
- and delivered at the speed required by rapidly growing digital infrastructure demand
In that context, clean energy economics are increasingly being shaped not by generation costs, but by system delivery constraints—particularly in regions experiencing rapid data center expansion.
The result is a grid where the limiting factor is shifting from power production to infrastructure timing and deliverability.