IEA: U.S. Data Center Electricity Use to Surge 130% by 2030
- AI Energy Demand Analysis Data Centers
Marin Holt
- 0
- 7 minutes read
- IEA projects U.S. data center electricity use rises 130% by 2030, adding 240 TWh
- Global data center consumption: 415 TWh in 2024 to 945 TWh by 2030
- AI accelerated servers growing 30% annually, four times faster than any other demand sector
- U.S. and China together account for nearly 80% of projected global growth
- Data center demand grows 15% per year through 2030, outpacing all other electricity sectors
U.S. data centers consumed approximately 415 TWh of electricity in 2024, representing 1.5% of global electricity use. The International Energy Agency projects that figure doubles to roughly 945 TWh globally by 2030, with the United States absorbing 240 TWh of that increase, a 130% rise. The driver is AI infrastructure: accelerated servers running machine learning workloads are growing at 30% annually, four times faster than any other sector of electricity demand.
What the Numbers Show
The IEA base case projects global data center consumption reaching 945 TWh by 2030, up from 415 TWh in 2024, growing at roughly 15% per year. Accelerated servers, the GPU-heavy systems running AI training and inference, account for 50% of the net increase. Cooling and infrastructure account for another 20%. U.S. per-capita data center electricity consumption stood at 540 kWh in 2024 and is projected to exceed 1,200 kWh per capita by 2030. By comparison, China faces a 170% increase and Europe a 70% increase over the same period. In the IEA high-growth Lift-Off scenario, global data center consumption reaches 1,700 TWh by 2035, equivalent to 4.4% of projected global electricity demand.
Why This Matters for Grid Planning
A 130% increase in U.S. data center demand over six years is not a gradual load growth that utilities absorb through routine capacity planning. It represents a structural shift in where electricity flows. Utilities designed their distribution and transmission networks around residential and industrial loads with predictable daily and seasonal patterns. A hyperscale data center demanding 200 to 500 MW on a 24/7 basis with near-zero tolerance for outages does not fit that planning model. Power purchase agreements reflect the gap: Microsoft surpassed Amazon as the largest corporate clean power buyer globally, with 34.7 GW contracted as of late 2025. The scale of procurement reflects both the load growth projections and the difficulty of accessing grid power fast enough to match construction timelines. Several large-load customers now co-locate on-site generation and storage specifically to begin operations before transmission infrastructure upgrades are complete.
Implementation Implications
The IEA report identifies two inflection points that will determine whether the Lift-Off or High Efficiency scenario materializes. First, cooling technology: data centers running AI workloads generate heat densities that air cooling cannot handle at rack densities above roughly 20 kW per rack. Liquid cooling, direct chip cooling, and immersion systems all reduce total facility power by improving PUE, but adoption is uneven. Second, AI model efficiency: training and inference power requirements per task have been declining as model architectures improve. If that trend continues, the per-workload power draw drops even as total workload volume rises. The net effect determines whether the U.S. adds 240 TWh or more. For grid operators, both scenarios require significant transmission and distribution investment starting immediately, because infrastructure permitting and construction timelines run 5 to 10 years.
Grid Engineering Implications
A 130% U.S. data center demand increase concentrated in 200-500 MW hyperscale campuses imposes near-unity 24/7 loading on transmission equipment designed for cyclic demand patterns. This accelerates transformer aging beyond the cyclic loading assumptions in IEEE C57.91-2011 thermal models, shortening transformer life and increasing the rate of in-service failures on circuits serving large AI campuses.
Critical Perspective
The IEA projects a 130% rise in U.S. data center electricity use by 2030. This growth rate far outpaces that of established tech giants like Amazon Web Services, which saw its AWS data center energy consumption increase by roughly 30% between 2022 and 2023. Are grid operators truly prepared for this scale of concentrated, 24/7 demand without substantial, and potentially costly, infrastructure overhauls? The forecast is credible on aggregate; the question is whether the 5 to 10 year transmission build-out timeline can keep pace with the 18-month corporate AI capex cycle.
Source
- IEA: Energy and AI: Energy Demand from AI (https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai)
