Key Takeaways

  • Global data-center electricity use is projected to more than double from 2024 to 2030, led by AI workloads.
  • Utilities face pressure to expand generation, transmission, distribution, and interconnection capacity while protecting other customers.
  • Microsoft, Google, and Amazon Web Services are increasingly treating access to power as a core infrastructure and location decision.

Data centers are rapidly becoming one of the defining electricity challenges of the decade. Global facilities consumed about 415 TWh in 2024, equivalent to 1.5% of worldwide electricity use. That figure could reach roughly 945 TWh by 2030, with artificial intelligence serving as the principal growth driver, according to the International Energy Agency.

The geographic concentration makes the challenge more acute. U.S. facilities represented approximately 45% of global data-center electricity demand in 2024, while China accounted for about 25% and Europe for 15%. This is not simply a question of producing more electricity globally. Capacity has to be available in particular regions, connected to specific campuses, and delivered reliably around the clock.

In the United States, data centers used an estimated 192 TWh in 2024, or 4.7% of national electricity consumption. The reference case could reach 649 TWh, equivalent to 11.8% of U.S. electricity use, by 2030, according to the Lawrence Berkeley National Laboratory. Data centers are also projected to represent nearly half of total U.S. electricity-demand growth through the end of the decade.

Aggregate energy figures only tell part of the story. AI campuses can introduce very large loads in one location, sometimes faster than local substations, transmission lines, and generation projects can be completed. A region might have adequate annual electricity production but still lack the infrastructure to connect another hyperscale development without creating congestion or reliability concerns.

AI-focused facilities are widening that gap. Their electricity consumption grew 50% in 2025, compared with 17% growth for total data-center demand. AI-focused consumption is projected to triple between 2025 and 2030. Training advanced models contributes to demand, but so does inference, the repeated processing involved when millions of users and enterprise applications call AI services throughout the day.

For utilities, the response extends well beyond adding power plants. New transmission corridors, distribution equipment, transformers, substations, storage systems, and grid-management capabilities can all be involved. Interconnection reform matters too. Projects can spend years waiting for studies and network upgrades, while data-center developers often work on much shorter construction schedules. Who pays when a new campus triggers an expensive regional upgrade? Regulators, utilities, developers, and existing ratepayers are still working through that question.

The generation mix is evolving at the same time. The IEA projects that renewable generation serving data centers will increase by more than 450 TWh by 2035. Wind and solar can provide substantial energy, although their variability means utilities and operators may also consider storage, flexible demand, firm generation, and on-site resources. Utility Dive has highlighted the broader industry debate over meeting AI-driven demand while balancing reliability, cost, and system planning.

Microsoft, Google, and Amazon Web Services are already treating power procurement as a strategic constraint rather than a routine facilities expense. Their hyperscale AI and cloud campuses are accelerating the pursuit of grid connections, renewable-energy contracts, and on-site generation. That changes site selection. Access to fiber and land still matters, but a credible route to hundreds of megawatts can determine whether a project proceeds.

Efficiency remains one lever, though not a complete answer. Power Usage Effectiveness, or PUE, helps operators compare total facility energy with the electricity used by computing equipment. Better cooling, higher server utilization, and more efficient chips can reduce overhead. Yet efficiency gains may be overtaken by rapid growth in computing volume, a familiar rebound effect in technology markets.

That said, the boom creates an opening for closer coordination. Utilities can use phased interconnection agreements and clearer cost-allocation rules, while operators can provide better load forecasts and some flexibility around noncritical computing. The companies that secure dependable electricity without shifting disproportionate costs onto surrounding communities are positioned to scale their infrastructure sustainably. In AI infrastructure, access to compute increasingly starts with access to the grid.