Yin WANG

Abstract

Artificial intelligence (AI) has become a central topic in contemporary competition policy debates. Yet a fundamental question remains unresolved: what should constitute the relevant unit of antitrust analysis in AI markets? Existing studies often approach AI-related competition issues from isolated technological components or specific market segments, resulting in fragmented analyses. Such approaches overlook that AI is not a single algorithm but is embedded in a complex value chain composed of multiple interrelated stages, where value creation and competitive advantages accumulate and reinforce one another across the AI lifecycle. Drawing on the OECD AI lifecycle framework, this paper develops a lifecycle value-chain approach to competition analysis. The framework conceptualizes AI as a multi-layered value chain consisting of four stages. By examining how market power may emerge, accumulate, and propagate across these stages, the framework identifies key sources of dominance within AI. Building on this perspective, the paper compares several technological paradigms, including generative AI, symbolic AI, agentic AI, and traditional machine learning, and highlights the heterogeneous mechanisms through which competitive advantages arise across these systems. By shifting the analytical focus from individual AI components to the broader system lifecycle, the proposed framework provides a structured approach for competition authorities to assess market power and potential anticompetitive conduct in emerging AI ecosystems.