Giacomo Santangelo

Abstract:
This paper examines the externality–equity nexus of artificial intelligence (AI) infrastructure. Using spatially weighted regression analysis on U.S. ZIP code data, I show that environmental burdens are disproportionately concentrated in lower-income communities while productivity benefits are captured by higher-income areas. These distributional asymmetries transform a traditional market failure into a justice failure, amplifying socioeconomic inequality. Methodologically, I integrate environmental vulnerability indices with measures of facility density and localized productivity spillovers, identifying both regressive externalities and progressive gains. My findings highlight the need for policies that address not only efficiency but also fairness in the governance of AI infrastructure.