Assessing the Feasibility of Federated Learning Deployment in Multi-Cloud AI Ecosystems

Abstract

The rise of data sovereignty concerns and privacy regulations has prompted a shift toward decentralized machine learning models. Federated Learning (FL), with its promise of data locality and collaborative model training, is a pivotal innovation. This paper examines the feasibility of deploying FL in multi-cloud AI ecosystems, focusing on infrastructure heterogeneity, data residency policies, interoperability, and performance benchmarks. We assess support across major cloud providers and model types, identifying technical barriers and strategic enablers for successful FL adoption.