Estimating GHG emissions from cloud computing: sources of inaccuracy, opportunities and challenges in location-based and use-based approaches

Allocating cloud computing greenhouse gas (GHG) emissions is essential for distributing responsibility for climate damages and informing strategies to reduce data center energy consumption and Scope 2 emissions. However, few studies investigated the practical usability of common GHG accounting frameworks for estimating cloud emissions at the national level. This study characterizes four sources of inaccuracies in estimating cloud-related GHG emissions and proposes three targeted interventions to address accounting risks. This work identifies estimation risks for Scopes 2 and 3 emissions when scaling organizational emission inventories of cloud consumers (carbon importers) and cloud hosts (carbon exporters) to national emission inventories. Current practices, which assign emissions based solely on the physical location of the emitting source, e.g. data centers, fail to account for the geographical separation between cloud operation and use. This may lead to an underestimation of total cloud-related GHG emissions and a disproportionate allocation of these emissions to carbon exporters. To mitigate these risks, this study introduces a use-based emissions attribution model, which allocates emissions based on cloud service consumption patterns and operational activities. The study also outlines three specific policy interventions to implement this approach: (i) stricter emission accounting rules, (ii) eco-labeling, and (iii) carbon border adjustment.