Data centre energy demand projections within shared socioeconomic pathways

Fan, Y.V., Wilson, C. ORCID: https://orcid.org/0000-0001-8164-3566, Kamiya, G., & Mastrucci, A. ORCID: https://orcid.org/0000-0002-5611-7780 (2026). Data centre energy demand projections within shared socioeconomic pathways. Energy and Climate Change 7 e100253. 10.1016/j.egycc.2026.100253.

Preview

Text 1-s2.0-S266627872600019X-main.pdf - Published Version Available under License Creative Commons Attribution. Download (2MB) | Preview

Abstract

Widespread digital transformation and the rapid expansion of data centres for artificial intelligence (AI) pose growing challenges for electricity systems and global climate targets. Many studies project near-term data centre energy demand, but few extend beyond five years due to the high degree of uncertainty in technological trends and a lack of robust and widely accepted methodologies. This is a problem for climate mitigation research that covers plausible scenario envelopes over longer time horizons to assess carbon budgets and warming outcomes. In this study, we extend data centre energy demand projections to 2050 within the Shared Socioeconomic Pathways (SSP) framework, aligning with common approaches in climate scenario analysis. Our method draws on three main historical periods since 2010 marked by distinctive trends in data centre service demand growth or efficiency gains, combined with projected future trends in digital transformation levels. Across the five SSP scenarios, we find global data centre electricity demand ranges widely, reaching between 1,800 and 5,000 TWh by 2050. We also find that the average weighted carbon intensity of electricity consumed by data centres must fall to below 100 gCO₂/kWh from 2030 to remain compatible with 1.5°C climate targets. Unlike most SSP-based studies, we find that uncertainties in data centre energy demand growth as a result of within-sector development and efficiency dynamics are wider than uncertainties from the different socioeconomic development trajectories captured by variation across SSP storylines. Rather than subsuming data centres within the commercial buildings sector as currently, we argue that data centres should be differentiated as a separate end-use sector in global energy statistics and modelling to account for its unique growth characteristics and uncertainties. Our findings provide a structured, scenario-based extension of anticipated near-term data centre energy needs. Our characterisation of uncertainties informs strategies for managing long-term digital infrastructure growth in line with global climate goals.

Actions (login required)

View Item