Artificial Intelligence (AI) is increasingly promoted as a catalyst for the Circular Economy (CE), enabling resource-efficient production, predictive maintenance, sustainable product-service systems, and closing material loops. However, the resource and energy demands of AI systems themselves, especially with the rise of large-scale models, raise concerns about their compatibility with CE principles. This study critically examines the circularity of AI, shifting the focus from AI as an enabler of the CE to AI as an object of CE strategies. Through a narrative literature review, we synthesize fragmented insights pertaining to AI and circularity. We identify significant knowledge gaps, particularly regarding hardware reuse, software sustainability, and the absence of tailored LCA methodologies for AI. Our analysis maps direct and indirect effects of AI on circularity and resource use. We abduct the concept of the Infosphere, i.e., the digital layer of society, to complement the established biosphere and technosphere dichotomy, emphasizing that AI’s impacts extend beyond hardware infrastructure. The introduction of the Infosphere should spark discussions about rethinking the role of software and AI as subjects in system-wide sustainability assessments. By exploring a first typology of the CE of AI, this paper lays a conceptual foundation for future research and policymaking, advocating AI systems that not only enable the transition to a CE but are also designed and governed in circular, sustainable ways.
Plociennik et al. (Thu,) studied this question.