IEA Warns of Doubling Data Center Energy Use Due to AI Boom

Subscribe to our free newsletter today to keep up to date with the latest energy news.

From generative AI to machine learning systems embedded in logistics, finance, and healthcare, the global AI ecosystem is expanding at a rapid pace. While its transformative potential draws significant attention, the unseen force powering AI’s evolution is the vast energy infrastructure behind it.

The International Energy Agency recently warned that electricity demand from data centers, which are central to AI’s operation, is on track to more than double by 2030. As AI becomes more data-intensive, particularly in language models and inference systems, the pressure on global power networks continues to grow.

Projected energy demands

This dramatic surge is driven by generative AI workloads, especially during model training and deployment. For example, training a single large language model such as GPT-4 can require tens of gigawatt-hours of electricity. Each time these AI systems are queried or deployed at scale, they depend on infrastructure operating continuously and often near full capacity.

Hyperscale data centers, operated by companies like Google, Microsoft, and Amazon, are rapidly expanding their physical footprint. In Northern Virginia, data center space now covers an area equal to eight Empire State Buildings. This expansion illustrates the urgency of developing energy-efficient infrastructure alongside technological innovation.

Increased electricity use is tied directly to emissions, particularly in regions where power grids still depend on fossil fuels. According to the International Monetary Fund, AI-related energy use could add as much as 1.7 gigatons of carbon dioxide emissions globally between 2025 and 2030.

The environmental impact extends beyond emissions. Concentrated data center operations can strain local infrastructure through thermal pollution, water use for cooling, and increased load on regional grids. A single data center may consume as much electricity as 100,000 homes, raising concerns about how these facilities fit into broader climate goals.

Environmental groups and some energy experts argue that without stronger accountability and investment in greener technologies, the environmental cost of AI could undermine its broader societal benefits. Regulators in some regions are pushing for mandatory energy reporting and transparency in emissions data from technology companies.

Meeting the power needs

Addressing AI’s energy demands involves more than producing electricity. The distribution and reliability of that power are equally critical. Many urban grids were not designed to accommodate 24/7 industrial-scale operations. As AI workloads increase, so does the risk of local grid failures, particularly during peak demand periods.

In deregulated energy markets such as the US, large data centers often secure discounted long-term power contracts, which can lead to market distortions. These practices may increase costs for residential users while putting pressure on energy providers to meet growing demand. Developing economies that aim to attract tech investment could find their public power systems overwhelmed.

Modernizing energy infrastructure requires large-scale investment. Grid upgrades, expanded generation capacity, and the addition of backup systems can take years to implement and demand coordination between public and private sectors.

Several mitigation strategies are under development. Some tech companies are co-locating data centers near renewable energy sources, such as hydroelectric dams or wind farms, to reduce transmission losses and dependence on fossil fuels.

AI also plays a role in optimizing energy efficiency. Applications in smart grid management, predictive maintenance, and demand forecasting are helping utilities reduce waste and improve system performance. In this sense, AI’s growing energy needs could also drive smarter energy use.

Technology firms are experimenting with advanced cooling systems, modular data center designs, and improved workload scheduling to enhance operational efficiency. Policymakers are introducing incentives and regulations to support cleaner AI operations. The European Union is currently evaluating rules that would set benchmarks for energy performance in AI-related infrastructure.

Sources:

• The Wall Street Journal
• International Energy Agency