Harnessing Large Load Growth

AI’s surge is no longer a future scenario; it’s already reshaping the U.S. power system. EPRI’s updated “Powering Intelligence” study shows just how consequential that shift could be. Data centers, particularly AI-focused facilities, are emerging as the largest driver of electricity demand growth in decades. By 2030, data centers could account for nine percent to 17 percent of total U.S. electricity demand, more than double today’s share.

After 2030, this dynamic is expected to shift, driven by expanding electrification across multiple sectors. By the mid-2030s to 2040, electric vehicle (EV) load is expected to surpass data center demand in most planning scenarios.

In a medium-growth projection, demand from EVs is projected to exceed data center load by 2036. Even in a high data center growth case, EV load surpasses it by 2041, reaching around 1,800 terawatt hours by 2050 compared to roughly 900 terawatt hours for data centers.

EVs represent a longer-term transformation, but despite their differences, both data centers and EVs share important similarities. Each represents a rapidly growing and uncertain source of load, driven by evolving technology and market dynamics.

Data center demand depends heavily on the trajectory of AI and computing efficiency, while EV demand hinges on adoption rates, policy incentives, and vehicle efficiency improvements. Data centers are driving a near-term demand spike, while EVs underpin sustained long-term load growth.

Electric Transportation Will Take the Lead

In the decades ahead, the electrification of transportation could become the dominant force shaping the grid. In most parts of the country, economic factors are accelerating this transition. EVs typically offer a lower cost per mile than gasoline vehicles across all states, with particularly strong savings where gasoline prices are high.

For medium- and heavy-duty vehicles, these savings can range from 35 percent to 82 percent, reinforcing the business case for fleet electrification.  In addition, electricity prices tend to be more stable than gasoline, which is subject to global oil market volatility.  This stability gives EV drivers and fleet operators more predictable operating costs.

Together, these trends are expected to accelerate EV adoption. For utilities, that means planning for a future where millions of vehicles collectively reshape electricity demand. Unlike data centers, which require large but localized infrastructure investments, EVs may likely drive widespread upgrades at the distribution level and require more granular planning.

The emerging picture is one of convergence: two powerful electrification trends reshaping the grid in different ways. The result is a fundamental shift in how the grid grows, positioning EVs as a central force in the next era of U.S. electricity demand. Together, data centers and EVs are increasing both the scale and complexity of demand on the power system. Both will play critical roles, increasing the urgency of grid planning and investment.

DCFlex’s Headroom and Flex MOSAIC™ Frameworks

Large loads materially affect power system operations, but they can also be a powerful tool to capitalize on the excess capacity available within the grid. These loads can be designed and operated to support reliability, manage congestion, and provide support during periods of system stress — whether by shifting workloads, modulating demand, or using onsite generation or storage.

However, inconsistencies across regions and markets have historically made it difficult for large load customers to understand which flexibility characteristics are most valuable to the grid.

Through its DCFlex initiative, EPRI is developing two complementary tools, the Headroom framework and Flex MOSAIC™, to help address these challenges and reshape how grid planners approach load growth. These frameworks are broadly applicable across electrification trends, including data centers, EV charging, and manufacturing.

EPRI’s Headroom framework will be designed as a practical, planner-oriented playbook to support near-term interconnection decisions and inform longer-term system investments. It helps electric customers and planners better understand how to make fuller use of existing grid capacity while maintaining reliability and supporting affordability.

The framework can provide realistic headroom estimates that reflect real-world operational complexity, evaluating opportunities without assuming immediate expansion of central generation, transmission, or energy storage.

To refine and validate this approach, EPRI is conducting a series of headroom case studies across diverse power systems in North America and Europe in the coming months. These analyses consider varying resource portfolios, market designs, regulatory contexts, and operational challenges, with the goal of producing a version of the framework that reflects industry consensus and can guide large-scale implementation.

These results can help electric companies provide clearer, more consistent signals to data centers and other large load developers, reducing fragmented approaches to incorporating flexibility into planning and operations.

Complementing this work, EPRI’s Flex MOSAIC initiative defines several classes of large load flexibility, and the specific grid needs each class can support. By establishing a common set of terms and transparent performance criteria, the voluntary framework provides a technical foundation that jurisdictions, markets, and system operators can adapt to regional requirements.

By aligning large load developers with grid operators, Flex MOSAIC can be a pathway to faster, more consistent, and more informed planning decisions that support more efficient grid integration.

For more information

• Electric Transportation
• Research Programs
• E-Roadmap
• Headroom Framework
• Flex MOSAIC™