Comparing carbon emissions of SaaS features across AWS, Azure, and GCP

Abstract

This thesis investigates the carbon emissions associated with the functionally equivalent Software-as-a-Service (SaaS) features offered by Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP). Through controlled experiments, we evaluated the emissions of three widely used services; document processing, language translation, and image analysis, using both cloud-native sustainability dashboards and a runtime-based proxy method. Our analysis revealed that emissions vary across providers for similar SaaS features. Native dashboards lack API-level granularity, making it essential to use a proxy-based approach. Based on our findings, we offer practical recommendations for cloud consumers to make informed decisions. The study is limited by the absence of realtime energy data, but it contributes to the broader goal of sustainable cloud adoption by identifying emissions of selected SaaS features and transparency gaps. Future work should focus on improving precision through direct energy measurement.