Key messages

  • AI is driving increased electricity demand, although the extent of its future impact on global energy demand is uncertain.
  • To simulate increased energy consumption due to AI in En-ROADS, reduce the energy efficiency of buildings and industry.


Background

The rapid growth of artificial intelligence (AI) is having a significant impact on the energy consumption and greenhouse gas emissions of data centers.

Energy demand in En-ROADS does not currently include estimates of future energy consumption from the growth of AI. Energy demand in the En-ROADS Baseline Scenario is compared against IEA World Energy Outlook (2023), NGFS (2023), and SSP Database Version 2.0 (2018). The scenarios and projections from these sources largely predate the mass adoption of ChatGPT and other AI technologies.

In 2024, data centers accounted for around 1% of global electricity consumption, according to the IEA. By 2026, electricity consumption from data centres, AI, and the cryptocurrency sector could double. After globally consuming an estimated 460 terawatt-hours (TWh) in 2022, data centres’ total electricity consumption could reach 1,050 TWh (3.8 exajoules) in 2026, an increase of 590 TWh (or 2.1 exajoules). For reference, 1,050 TWh is roughly equivalent to the annual electricity consumption of Japan.

According to Lawrence Berkeley National Laboratory, data centers in the U.S. could account for up to 12% of all U.S. electricity consumption by 2028, or 132 gigawatts annually, triple what they consumed in 2023.

While many data center operators are planning to meet increased electricity demand with renewable energy (e.g., Google), others are turning to nuclear power (e.g., Microsoft).

How do I simulate the effect of increased energy demand due to AI in En-ROADS?

We expect to be able to update the model to reflect updated forecasts in 2025. In the meantime, the primary way to simulate the impact of increased energy demand is by reducing the energy efficiency of buildings and industry. This has the impact of increasing final energy consumption. To make this change take effect sooner, also reduce the “Years to achieve energy efficiency improvement rates” slider under Simulation > Assumptions > Energy > Policy timing.

See an example scenario in En-ROADS.

While data center energy demand is growing, it's important to note that both data center energy efficiency (measured by power usage effectiveness) and AI computational power continue to improve. AI itself is being deployed to further improve energy efficiency and energy use in buildings and industry, all of which may have some compensating effect.


Further reading

Generative AI – The Power and the Glory: A useful overview of the energy demand of AI data centers, from BloombergNEF.