How Are Data Centers Affecting the U.S. Power Grid?

After nearly 15 years of relatively flat electricity demand, the U.S. is seeing a steady increase in power. The U.S. Energy Information Administration (EIA) says electricity demand has grown by an average of 2.1% per year over the last five years. It projects continued growth through 2050, with data center server energy use as a major factor.

At the same time, EPRI estimates data centers now account for 4% to 5% of total U.S. electricity demand. This demand could reach 9% to 17% by 2030. That is a meaningful shift for developers, utilities, and enterprise leaders alike.

This shift is influencing how utilities plan, how developers build, and how organizations approach growth. Understanding where things stand today and where they’re headed can help make sense of what’s happening across the data center landscape. Let’s get into it.

Who Powers Data Centers in the U.S.?

The broader electric grid powers most data centers in the U.S. Their electricity comes from a mix of regional grid operators, utilities, generators, and transmission providers rather than from a single source. Some of the energy sources these providers are using include natural gas, wind, solar, nuclear, and onsite battery storage.

Depending on a facility’s location, this power may be coordinated through organizations such as PJM Interconnection, ERCOT, MISO, CAISO, SPP, or NYISO, and then delivered by local utilities and other energy providers. Data center power is shaped as much by regional grid conditions as it is by facility design.

How Data Center Power Is Delivered Across the Grid

This system wasn’t originally designed with today’s demand profile in mind. For years, utilities planned around modest, predictable growth. But large-scale data centers—especially hyperscale and AI-focused facilities—can now come online at hundreds of megawatts and sometimes approach gigawatt scale. The U.S. power grid was built around 1% to 2% annual load growth, which looks very different from the concentrated demand increases happening today. As a result, power availability is becoming an active part of development strategy rather than a given input.

Across regions, this shift is playing out in different ways depending on local infrastructure, regulation, and market structure. In markets like PJM—the largest U.S. grid operator, which serves more than 67 million people across 13 states and D.C.—regulators and utilities are actively working through how to integrate large data center loads while maintaining reliability and balancing costs. PJM isn’t the only market facing these challenges. But it offers a clear example of how grid operators are adapting as large, fast-moving loads reshape demand patterns.

Even with Renewables, the Grid Still Matters

At the same time, the electricity used by data centers continues to reflect the broader U.S. energy mix. In 2025, the EIA reported that natural gas accounted for about 41% of utility-scale electricity generation, nuclear for about 18%, and wind and solar together reached a record 17%.

While many data center operators are increasing their use of renewable energy contracts, the physical electricity serving facilities still depends on the regional grid where they operate.

What Has Changed Because of Data Center Expansion

Data center growth has become a visible driver of electricity demand at both the regional and national level. Berkeley Lab estimates U.S. data centers consumed 176 terawatt-hours of electricity in 2023, up from 58 terawatt-hours in 2014. The lab projects that number could rise to 325 to 580 terawatt-hours by 2028. EPRI’s 2026 analysis shows that growth continuing, with demand potentially reaching 380 to 790 terawatt-hours by 2030.

Growth varies widely by facility type. A large new data center can represent 100 to 1,000 megawatts of load, roughly equivalent to the electricity demand of 80,000 to 800,000 average U.S. homes.

Smaller facilities operate at a very different scale, but the largest hyperscale projects are the ones driving today’s power conversations. When those projects cluster in a single region, their impact becomes much more visible.

AI has made that growth more immediate. AI workloads account for about 15% to 25% of data center electricity use today. And that share is rising as adoption expands. The International Energy Agency (IEA) reported in April 2026 that overall data center electricity demand grew 17% in 2025. And AI-focused sites are increasing even faster.

Even as efficiency improves, the scale of new compute workloads continues to push energy demand upward.

How Data Centers Are Affecting the Power Grid

At a national level, data centers still represent a relatively small share of electricity demand. But how does that demand show up at the regional level?

EPRI estimates data centers account for 4% to 5% of U.S. electricity demand overall. Yet in Virginia that share already exceeds 20% and could rise to 39% to 57% by 2030. Several other states could also cross the 20% threshold in the coming years.

Large Loads Are Reshaping Grid Planning

That concentration changes how the grid is planned and operated. Large, fast-growing loads can trigger new transmission requirements, additional generation capacity, and updated forecasting models. Utilities and regulators are also working through how those infrastructure costs should be allocated as demand grows. The Federal Energy Regulatory Commission’s (FERC) ongoing work around large-load interconnection reflects the need to integrate these projects in a way that remains reliable and fair for all customers.

Timing Is Becoming a Big Constraint

The dynamic is not just about volume. It is also about timing. Data centers can move from planning to operation relatively quickly, while the infrastructure required to support that growth often takes much longer to build. These timeline factors are becoming a defining feature of today’s power environment.

These regional pressures are reshaping how utilities plan and how developers prioritize sites. But they are also highlighting a big challenge for data center teams: how quickly power can be secured.

What Does the Next Decade Look Like for Data Centers?

This challenge is often described as time to power—the time required to secure reliable electricity, complete interconnection work, and bring supporting infrastructure online. It is quickly becoming one of the most important constraints on data center development.

The constraint is not driven by a single issue. It is the result of several factors converging at once. The IEA reports that supply chains for gas turbines, transformers, advanced chips, and other critical equipment have tightened. At the same time, planning and regulatory systems are also slowing grid connections and approvals. At the same time, the scale of proposed data center growth continues to expand.

Today, access to power is a key factor in site selection. Developers are prioritizing locations where capacity is available or where interconnection timelines are more predictable. Others are exploring options such as co-location, onsite generation, or more flexible load strategies to reduce reliance on the traditional grid process.

As timelines stretch, some hyperscalers are stepping into roles traditionally filled by utilities, securing or developing their own power solutions to avoid both delays and grid strain.

Turning Data Center Power Strategy Into Delivery

As power availability becomes more central to how projects move forward, data center power has shifted from a background consideration to a core factor in project planning. For developers and operators, the challenge is not only how much power is needed, but how quickly it can be secured and how reliably it can be delivered over time.

Organizations are rethinking how they plan and execute data center projects across technology, engineering, operations, and site development. Insight Global helps data center operators navigate that complexity by connecting the right talent and technical support across planning, engineering, and execution. That way, your projects move forward in a tighter power environment across the data center life cycle.