Physical Resource Coherence: The Missing Dimension of the AI Infrastructure Economy

Abstract The rapid expansion of artificial intelligence has generated unprecedented investment in computing infrastructure, semiconductor manufacturing, data centres, energy systems and supporting supply chains. Yet the sustainability of this expansion is often assessed through financial resources, technological performance and institutional arrangements without sufficiently examining the physical resources required to support continued deployment. This paper introduces Physical Resource Coherence (PRC) as the fifth proposition in the Governance Coherence (GC) research programme and as a fourth foundational dimension of systemic coherence, alongside Financial Coherence (FC), Technological Coherence (TC) and Institutional Coherence (IC). The central argument is that financial commitments, technological capability and institutional coordination cannot independently guarantee sustainable deployment when the physical resources supporting them are constrained, inaccessible, insufficiently adaptable or difficult to recover. The paper introduces the Physical Support Gap (PSG), a diagnostic construct describing the mismatch between physical resource requirements generated by deployment and the adaptive capacity available to meet those requirements. Two complementary measurement approaches are proposed: a growth-mismatch measure and a dominant-resource-pressure measure. The latter recognizes that aggregate resource availability can conceal critical bottlenecks that cannot readily be compensated for by surpluses elsewhere. The paper also examines the relationship between technological efficiency, demand expansion and the Jevons paradox, and connects physical infrastructure accumulation to Carlota Pérez's theory of technological revolutions. A proposed structural equation model (SEM) provides a basis for subsequent empirical investigation. The contribution is conceptual and methodological. The proposed relationships and measurements require operational definition, data collection and empirical validation. The paper does not claim that the proposed model has already established causal relationships.