Electricity demand projections have been sharply revised upward, raising risks to North American grid planning and reliability. NERC now forecasts 224 gigawatts of additional peak power capacity over the next ten years—69 percent higher than its projection from the previous year. This revision poses governance challenges: decarbonizing heating, transportation, and industry while also expanding digital infrastructure are competing for access to the same grid without a clear, uniform federal rule for allocating the costs of reinforcement tied to large loads.

The essentials

  • NERC forecasts 224 GW of additional peak power capacity across North America over ten years, with data centers identified as the primary source of growth, among other factors contributing to the gap with the 2024 projection (NERC, 2025 Long-Term Reliability Assessment).
  • The U.S. federal energy regulator, FERC, has ordered six regional operators to revise their connection rules for very large loads and is evaluating the need for more explicit rules tailored to very large loads.
  • The pressure mechanism: connection queues are filling with competing projects—solar parks, wind farms, industrial heat pumps, data centers—even as production queues are subject to readiness criteria and firm deadlines.
  • The positive tipping points of the energy transition (electric vehicles, heat pumps, green industry) depend on the same overloaded grid, which could affect their adoption pace.
  • The open question for the decade is whether the digital economy and decarbonization can coexist on the grid, or whether their competition requires explicit political arbitration.

224 GW in ten years: what lies behind a 69 percent revision

One gigawatt is enough to power roughly 750,000 American homes under normal conditions. NERC forecasts 224 additional gigawatts at summer peak by 2035. To understand what this gap represents relative to the 2024 projection, one must grasp how NERC builds its models: it aggregates connection declarations from regional operators, consumption projections from states, and declared investment plans. For twenty years, this method produced stable results. Demand grew slowly, reliability margins remained comfortable.

The boom in data centers increases the uncertainty and volatility of load forecasting. New-generation data centers, those running artificial intelligence models, consume between 100 and 500 megawatts each—as much as mid-size cities. Their developers file connection requests on timelines that grid operators had not anticipated. Result: connection queues remain very large and slow, yet their active national volume declined in 2024.

The 69 percent gap between the 2024 and 2025 projections does not reflect an error in earlier models. It reflects a real increase in demand from data center operators and other large loads. The revision is a governance signal as much as a demand signal.

Connection queues overflow in all regional markets

The North American electrical grid operates through regional markets managed by independent operators—PJM on the East Coast, MISO in the center, ERCOT in Texas, CAISO in California, among others. Each manages its own connection queue with its own rules about who pays what and within what timeframe.

These rules were designed for electricity producers—power plants, solar parks, wind farms—seeking to inject power onto the grid. They were not conceived for massive consumers seeking to connect at power levels previously reserved for large industries. Data centers belong to an emerging category of very large loads with novel characteristics, not an entirely nonexistent category of large loads.

In the PJM market, which covers thirteen states from Pennsylvania to Illinois, 212 active projects from Transition Cycle 1 represented approximately 14 GW in December 2024. Volume and speculative projects congested processing, leading PJM to adopt stricter readiness criteria. Solar and wind park operators complain that their projects wait years alongside speculative data center requests that monopolize the interconnection service’s study capacity.

FERC acknowledged this dysfunction in June 2026 by ordering six regional operators to propose new rules for integrating very large loads. The order calls for examining and improving application and study processes, funding rules, and network reinforcement cost allocation. Among pending requests are industrial decarbonization and electrification projects, which add to data center demand and jointly increase network connection needs.

The digital economy and decarbonization facing the same bottleneck

Tim Lenton, in his work on positive tipping points, has mapped how certain technologies—solar, electric vehicles, heat pumps—can self-amplify once they reach a critical adoption threshold. Cost curves collapse, behaviors change, policies follow. His book Positive Tipping Points documents several such dynamics underway since 2020.

These tipping points assume an electrical grid capable of absorbing them simultaneously, which poses capacity challenges. An electric vehicle needs only a garage outlet, but it needs current to charge. A heat pump replaces a gas boiler; it consumes electricity. A factory electrifying its furnaces replaces gas or coal; it adds megawatts to demand. These tipping points converge on the same grid at the same moment as digital demand.

The pressure is concrete. In several regions, solar park projects intended to power industrial electrification programs face prolonged connection delays. Regional operators cite saturation of transformation substations and transmission lines. Meanwhile, some data centers provisionally fed by other sources, awaiting grid connection, continue operating. The optimism of systemic tipping points collides with the inertia of shared infrastructure.

Daron Acemoglu, in his work on technology and power, questions the allocation of scarce resources and its beneficiaries. The electrical grid is public infrastructure whose reinforcement costs are largely mutualized among consumers through regulated tariffs. If data centers generate a significant share of reinforcement investments, existing tariffs may mutualize a large portion of these reinforcement costs, without the share actually paid by data centers being precisely established—raising a question of fair cost allocation. FERC is attempting to correct this situation with its order on connection rules.

Who pays for grid reinforcement: the question nobody has settled

The economics of electrical connection work on a simple principle in theory: whoever connects pays the costs they generate on the grid. In practice, this principle is applied very differently across regional markets. Some operators apply “deep cost allocation”: the newcomer fully finances the reinforcements their connection makes necessary. Others apply partial mutualization among all grid users.

For a 50-megawatt wind turbine, the difference is manageable. For a large data center campus, the difference can be very significant. Data center operators, particularly those of major technology platforms, have the financial capacity to absorb these costs. Several have signed long-term renewable electricity purchase agreements to finance new solar or wind parks. Microsoft, Google, and Amazon have announced investments in new civilian nuclear capacity to power their centers.

But between intention and physical grid connection lies the rulebook and the resulting delays. A solar park can face significant delays before being physically connected to the grid, during which the load it was supposed to feed draws power through other channels. This time gap, as much as the costs, poses problems.

Responses to the June 2026 orders are due to FERC in November 2026, after the initial 60-day period. Among proposed improvements are readiness mechanisms and financial commitments to filter speculative requests, as well as examination of possible prioritization according to infrastructure objectives.

Capacity the grid must absorb between 2026 and 2036

NERC’s decade-long projection is a planning document, not a certain forecast. The 224 GW corresponds to NERC’s summer peak growth forecast for 2025–2035. Corrections are likely. But pressure on the grid should remain significant in coming years.

Three trajectories are taking shape for the decade, without current data allowing any to be favored with certainty.

In the first, connection rules reformed by FERC succeed in disciplining queues, data centers finance a significant share of the reinforcements they generate, and decarbonization projects advance in parallel. This trajectory assumes rapid regulatory execution and cooperation from regional operators—two conditions that recent history of American grid regulation does not permit us to take for granted.

In the second, competition for connections slows both dynamics simultaneously. Data centers feed provisionally through off-grid contracts, including on-site gas generators. Industrial decarbonization and electric mobility projects fall behind in the most saturated regions. The positive tipping point documented by Lenton remains real but slower, geographically uneven, concentrated in states where the grid still has margins.

In the third, arbitration becomes explicitly political. States or Congress legislate on network access priority, as Virginia—the state home to the world’s largest concentration of data centers—has begun doing through transparency laws on consumption. This politicization can accelerate or block decisions depending on the coalitions in place.

The most instructive signal to track through 2027 will be the response of the six regional operators to FERC’s order: the quality of proposed mechanisms, their application timeline, and data center operators’ reaction to stricter reinforcement financing rules. These responses will indicate whether the American grid can manage dual pressure or must choose.

Signals that could break the deadlock

Several technical initiatives could ease constraints without awaiting complete regulatory reform. Active demand management, which lets data centers adjust consumption based on real-time price signals, is already being tested in several regional markets. A data center accepting a 20 percent consumption reduction during summer heat peaks frees capacity for other uses—a service to the grid, not a concession.

Large-scale energy storage, particularly high-power batteries installed upstream of saturated substations, can cushion demand peaks and defer costly reinforcement investments. Several regional operators have begun integrating these resources into their planning. Published articles on electrifying cities in extreme heat conditions show this grid pressure is simultaneous on multiple fronts: data centers are only one vector among others.

Geographic relocation of data centers to regions with grid margins is another avenue. Montana, Wyoming, and the Dakotas have underutilized transmission capacity and significant wind potential. Land costs and water availability for cooling complicate this option, but several projects are underway in these states.

Finally, transparency on actual data center consumption is a prerequisite for any serious arbitration. The geographic concentration of AI professions in a few metropolitan areas creates additional pressure on already-strained grid nodes, but this pressure is hard to quantify due to lack of public data on per-site consumption. Virginia legislated on this in 2024; other states are examining similar requirements.

Reforming connection rules is necessary, not sufficient. It does not resolve the question of total capacity to deploy, nor that of financing new high-voltage transmission lines whose authorization and construction timelines span a decade. Sharing a grid without explicit access rules amounts to letting the first arrival define conditions for everyone else.

Sources

  1. North American Electric Reliability Corporation, 2025 Long-Term Reliability Assessment, October 2025. https://prod.nerc.com/globalassets/our-work/assessments/nerc_ltra_2025.pdf
  2. Federal Energy Regulatory Commission, FERC Launches Aggressive Targeted Action to Speed Large Load Integration, June 2026. https://ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration
  3. Tim Lenton, Positive Tipping Points: How to Fix the Climate Crisis, Oxford University Press. https://global.oup.com/academic/product/positive-tipping-points-9780198937678
  4. Daron Acemoglu and Simon Johnson, Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity, PublicAffairs, 2023.
  5. PJM Interconnection, connection queue data, 2024 annual report (available at pjm.com, Planning & Interconnection section).