Global Steel Decarbonization under Policy, Technology, and Resource Constraints: A Process-Unit Agent-Based Model

Abstract Deep decarbonization of steel requires low-emission technologies to be deployed through a long-lived and heterogeneous asset fleet without disrupting steel supply. We developed a country-resolved agent-based model that follows global steel process units annually from 2023 to 2060. Units choose among continued operation, renewal, retrofit, conversion, retirement, and new construction, subject to construction time, linked upstream capacity, national production pathways, and annual supplies of scrap, hydrogen, natural gas, and pellets. Four scenarios distinguish stronger climate policy from improved technology and resource availability. Global process-chain CO2 emissions decline by 67–79% by 2060, reaching 0.476–0.737 Gt. The integrated case achieves the lowest emissions, but the two-component interventions do not independently outperform the reference case. Stronger policy changes when assets enter a finite transition queue, whereas technology and material conditions determine whether replacement routes can be built and operated at that time. By 2060, Scrap-EAF supplies 74–80% of crude steel, but residual unabated blast furnaces and coal-based direct reduction keep national steel pathways above their allocated levels in many countries. Deep decarbonization, therefore, depends on coordinating asset turnover, construction, and complementary material systems, rather than on technology availability or carbon-price pressure alone.