From 2010 to 2020, data center electricity consumption remained flat despite exponential data growth. However, since 2020, most industry forecast models now project significant growth in data center electricity consumption through 2030. This paper investigates the reasons behind this shift. We use a method that correlates annual information and compute technology (ICT) energy consumption with annual data production to frame the conversation between infrastructure and ICT technology providers. This correlation suggests a disruption in compute efficiency starting around 2020, with recovery projected to begin in 2026 and impact visible by 2030. The paper demonstrates that small improvements over currently planned projections in power usage effectiveness (PUE) and compute efficiency from 2026 can potentially "bend" the energy growth curve by 17%, decoupling data growth from compute energy consumption. This industry-led approach allows the world to prioritize the supply of clean energy. This focus fuels the economic and societal progress that comes with digitalization, rather than planning for seemingly unbounded data center energy growth.
Files
File Name | |
WP212_V2_EN.pdf
|
|