Summary: The rapid proliferation of Artificial Intelligence is fundamentally altering the trajectory of global data center energy consumption. According to the IEA, data centers accounted for approximately 415 terawatt-hours (TWh) in 2024—roughly 1.5% of global electricity demand. However, this figure is projected to double by 2030, reaching 945 TWh as "accelerated servers" (those equipped with high-performance GPUs for AI) grow at an annual rate of 30%.

While data centers will likely remain a small portion of total global demand (around 3% by 2030), their impact is intensified by extreme geographic concentration. The United States and China currently dominate this landscape; notably, U.S. per-capita data center energy use is an order of magnitude higher than other regions and is expected to reach 1,200 kWh by 2030. The report highlights three potential futures: a "Lift-Off" scenario where rapid AI adoption pushes demand above 1,700 TWh by 2035; a "High Efficiency" path where software and hardware breakthroughs mitigate growth; and a "Headwinds" scenario where supply chain constraints and slower adoption lead to a plateau. Ultimately, while AI is a significant driver, it remains one of several factors—alongside electric vehicles and industrial electrification—challenging the capacity and flexibility of modern power grids.

Source: https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai