IEA Projects Data Center Energy Demand to Double by 2030 Due to AI Surge

A recent report by the International Energy Agency (IEA) projects global electricity use from data centers to rise more than double to 945 terawatt-hours (TWh) annually by 2030 from 415TWh in 2024. The steep rise primarily results from the fast expansion of artificial intelligence (AI) applications that are fueling the energy consumption of data processing facilities. The United States and China are predicted to be the highest contributors to this growth, and they will account for almost 80 percent of the overall increase that is forecast. In the US alone, energy consumption in data centers will increase by 240TWh, a 130 percent rise, while China will reach 175TWh, a 170 percent surge from the current statistics. Other regions such as Europe and Japan will also see significant growth, with rises of 45TWh and 15TWh, respectively.

To satisfy this increasing demand, the energy mix for data centers will probably undergo a radical transformation. Renewable electricity generation, particularly from hydro, solar PV, and wind, is projected to grow more than 450TWh by 2035. Renewables now account for 27 percent of data center power supply, which is projected to grow to 50 percent by 2030. This growth in the percentage of renewable energy means that even with this increase, there will continue to be a demand for dispatchable sources of energy. Natural gas, which contributes 26 percent of current data center electricity usage, will grow by 175TWh, the majority of that concentrated in the United States. Nuclear energy, with a 15 percent market share, will also play a crucial role in ensuring energy dependability.

It indicates that this growing demand will require significant investment in transmission infrastructure. Investment shortages, e.g., up to 20 percent postponements of scheduled data center projects. Supply chains, particularly that for manufacturing generation equipment, could be an added challenge. The geographical location of the data centers would most likely shift. Now, over a majority of US data centers agglomerate in already formed clusters but can be shipped farther out with a siting, says the IEA, that frees up more room and thus reduces pressure on the grids. Building off-grid capabilities in energy-abundant locations such as Texas can also help enable this change.

Of more general concern is the effect of AI itself in generating and perhaps mitigating energy demand. AI-attuned data centers are power-intensive, yet AI-driven technologies possess the ability to improve the efficiency of the energy system. AI-driven applications, for instance, may be used to make electricity generation, transmission, and usage more optimal. But the energy sector hitherto has not been successful in leveraging the potential of AI due to limited access to data, digital infrastructure, and cyber security concerns. The report calls for more investment in digitalization and employee skills to unlock these potential benefits.

From a supply of materials perspective, the expansion of AI data centers will place global resources under strain. Demand for gallium, a metal used in semiconductors and power electronics, is forecast to exceed 10 percent of available supply by 2030. As China produces 99 percent of the world's refined gallium, any disruption to its supply chain could lead to enormous delays for data center developers globally.

From a green point of view, the IEA predicts that carbon emissions from data center use of electricity may rise from 180 million tons (Mt) now to 300Mt in the base case and up to 500Mt in the high-growth scenario by 2035. However, the report is quick to note that the prolific use of AI to save energy and reduce emissions has the potential to wipe out such increases. If adopted in large numbers, these applications can lead to emissions reductions of five percent of global energy-related emissions by 2035, sparing AI the likely difference in climate wars.

Source/Credits:
Zachary Skidmore, April 11, 2025