How Much More Power Can the U.S. Grid Provide for AI?

The anticipated growth in electricity demand from artificial intelligence (AI) is large, rapid, and geographically concentrated. Because of uncertainty about which planned generation projects will be completed, it is difficult to assess whether future U.S. grid capacity will keep pace with demand from AI data centers. An additional challenge lies in translating announced nameplate capacity into comparable estimates of reliable power that can meet large, inflexible loads to support data center power needs.^[1] These factors complicate efforts to estimate how much additional power capacity the United States is likely to have by 2030.

Estimating Additional Power Capacity

A team of RAND researchers estimated how much additional power capacity the United States is likely to have by 2030 by translating planned electricity supply resources into a common measure of reliable capacity. For front-of-the-meter (FTM) resources,^[2] the researchers analyzed planned generation and storage projects based on independent system operator (ISO) interconnection queues and federal generation data and applied historical completion rates by region and technology.^[3] This process adjusted for resource contributions to reliability and accounted for planned retirements to estimate net additions to grid capacity.

For behind-the-meter (BTM) resources,^[4] such as customer-sited solar and battery storage, the researchers relied on national deployment projections through 2030 and converted projected nameplate capacity into effective capacity using the same framework as in FTM.^[5] Using this approach, the researchers estimated how much BTM resources could reduce peak grid demand and free up capacity for large loads, such as AI data centers.

Estimates of Front-of-the-Meter and Behind-the-Meter Net Availability Capacity

The researchers found that the United States could add approximately 82 gigawatts (GW) of net available capacity by 2030,^[6] consisting of 33 GW from FTM resources and 49 GW from BTM resources; see Table 1.

Figure 1 illustrates how planned FTM generation and storage projects are translated from 1,086 GW of nameplate capacity into the estimated 33 GW of net available capacity after project attrition, retirements of existing plants, and differences in resource contributions to reliability are accounted for.

Interconnection queue funnel: 1,086 GW shrinks to 33 GW net available capacity after losses.

Step 1. Estimate planned energy resources and consider deliverability by aggregating nameplate capacity that is planned with a reliable connection type. Net available capacity (data center power need) = 1,086 gigawatts. Projects that do not have reliable deliverability status = -71 gigawatts.

Step 2. Estimate project completion by applying historical completion rates to account for withdrawals. Net available capacity (data center power need) = 1,015 GW gigawatts. Projects that fail to reach completion = -864 gigawatts.

Step 3. Account for resource- and region-specific performance by using CAF and T&D loss. Net available capacity (data center power need) = 151 gigawatts. Portion of capacities and energy that are not dependable to serve additional load = -74 gigawatts.

Step 4. Analyze net additions by considering planned retirement. Net available capacity (data center power need) = 77 gigawatts. Capacity unavailable because of retirement = -44 gigawatts.

NOTE: CAF = capacity accreditation factor; T&D = transmission and distribution. Values in gigawatts might not sum precisely because of rounding.

^a The analysis includes only projects with interconnection types that ensure a reliable power supply to data centers, overcoming transmission line constraints, available grid capacity, and potential congestion. Further details are available in the accompanying research report.

Net FTM capacity additions vary substantially by region, reflecting differences in generation mixes, the composition of interconnection queues, and retirement schedules. Growth in ERCOT interconnection requests drives a large share of forecasted additions (see Table 2), while queue requests are paused in CAISO and PJM.^[7] Figure 2 provides a consolidated view of how different resource types contribute to the gap between planned nameplate capacity additions and net available capacity.

Resource	Planned Nameplate Capacity	Expected Nameplate Capacity	Available Capacity Additions	Planned Retirement	Net Available Capacity
Coal	0	0	0	-31	-31
Gas	64	19	14	-12	2
Other	0	0	0	0	0
Nuclear	0	0	0	0	0
Biomass	0	0	0	0	0
Solar	314	42	5	0	5
Wind	114	19	4	0	4
Geothermal	0	0	0	0	0
Hydro	2	1	0	0	0
Hybrid	68	5	3	0	3
Storage	451	65	50	0	50
Total	1015 GW	151 GW	77 GW	-44 GW	33 GW

SOURCES: Planned nameplate capacity is estimated from October 2024 interconnection queues for CAISO, NY-ISO, ERCOT, NE-ISO, SPP, and MISO, and non-ISO planned capacity is as reported in U.S. Energy Information Administration, 2025. Expected nameplate capacity applies historical completion rates to planned capacity, as estimated from CAISO, NY-ISO, SPP, and MISO interconnection queues and self-reported by ERCOT. Non-ISO and NE-ISO completion rates are estimated as averages of completion rates across other regions. Available capacity, then, also takes transmission line losses and effective load-carrying capability (ELCC) estimates for CAISO, NY-ISO, ERCOT, NE-ISO, SPP, and MISO into account, using an average for non-ISO resources. Planned retirement resources are estimated by applying ELCC and transmission loss factors to nameplate capacity of resources scheduled for retirement in 2025–2030 that appear in U.S. Energy Information Administration, 2025. Net available capacity for 2025–2030 is equal to available capacity additions minus planned retirements.

BTM resources play an important role in reducing peak grid demand and potentially freeing up FTM capacity. Projected BTM additions total 49 GW by 2030 (Table 3), though the contribution varies by region. In some regions, accounting for BTM resources materially changes the aggregate capacity outlook, while in others the effect is modest.

Key Limitations and Assumptions

The study had some important limitations and assumptions. The researchers assumed that data center load profiles are firm and inflexible. They also assumed static estimates of resource contributions to reliability. For BTM resources, the researchers based their estimates on aggregate projections rather than public project announcements. They also assumed that BTM resources will continue to be built out and that retirements will be limited.

Recommendations and Policy Implications

The study's findings suggest that planned power capacity expansions are unlikely to align evenly with the geographic distribution and timing of rapidly growing electricity demand, particularly from large, inflexible loads, such as AI data centers. Policymakers and planners may therefore need to focus explicitly on locational adequacy, especially in regions where net additions of reliable capacity are limited. Incorporating realistic project completion rates, retirements, and resource reliability is critical for assessing future grid adequacy.

The analysis also highlights the importance of managing uncertainty through diversified resource portfolios. Although BTM resources can meaningfully reduce peak demand in some regions, they are unlikely to fully offset shortfalls in grid-connected supply where constraints are most severe.

Further research could refine these estimates by incorporating transmission expansion scenarios, more-granular locational data, and alternative demand growth trajectories from uncertain large-load growth.

Notes

1. Nameplate capacity is the maximum theoretical output of a power-generation facility as specified by the manufacturer.Return to content ⤴
2. Front of the meter refers to energy systems that are connected to the utility grid and supply electricity to a broad area, typically including large-scale generation and storage facilities.Return to content ⤴
3. ISOs operate and run the regional electric transmission grids, ensuring transmission access, planning, operation, and dispatch of energy.Return to content ⤴
4. Behind the meter refers to energy systems located on the customer's side of the utility meter. These systems generate, store, or manage electricity without passing through the meter that measures power consumption from the grid.Return to content ⤴
5. Effective capacity is a technology-agnostic reliable capacity. It is determined by capacity accreditation factors and used by energy planners for resource adequacy assessment.Return to content ⤴
6. Available capacity refers to the researchers' estimate of future, deliverable, and aggregated power. It is estimated from active and anticipated energy projects seeking interconnection to the grid and builds on effective capacity by adding completion rate, losses, and deliverability. More information can be found in the glossary of the accompanying research report, Assessing the United States' Additional AI Power Capacity by 2030: Estimating Short-Term Increases in Electricity Generation and the Ability to Meet Growth in Power Demand.Return to content ⤴
7. ERCOT, CAISO, and PJM are all ISOs.Return to content ⤴

Available for Download

Download PDF

Topics

• Artificial Intelligence
• Energy Consumption
• Energy Distribution
• Energy Resources