This review article examines the current state and projected trajectory of micro-metallurgical production, defined as small and mid-scale secondary steelmaking operations based on scrap feedstock, localized demand, and energy-efficient rolling and melting technologies, within the global steel system. The article synthesizes quantitative market data, regulatory developments, and the rapidly maturing application of artificial intelligence to metallurgical processes, drawing on peer-reviewed literature, industry analyses, and primary data from 2024 through 2026. Three structural forces are reshaping the sector simultaneously: the global shift from integrated blast-furnace production toward electric arc furnace (EAF) and direct-reduced-iron (DRI) routes, driven by scrap availability and decarbonization policy including the European Union Carbon Border Adjustment Mechanism (CBAM); the progressive lowering of minimum efficient scale in secondary steelmaking through induction heating, cross-wedge rolling, and modular plant design, which makes facilities of 20,000 to 350,000 tonnes per year commercially viable; and the integration of artificial intelligence, including machine learning, digital twins, computer vision, and generative models, into furnace control, predictive maintenance, and scrap characterization. The article argues that these three forces are mutually reinforcing, and that micro-metallurgical facilities in emerging mining-intensive economies represent the most likely mechanism through which regional demand for specialized steel products, particularly grinding media, rolling stock components, and wear-resistant consumables, will be served through the 2030s. The article identifies five research and policy priorities: standardization of AI-ready data architectures for small-scale plants, development of physics-informed machine learning models suited to induction-based production, regulatory frameworks that extend decarbonization incentives beyond integrated producers, workforce development that bridges metallurgy and data science, and empirical validation of the financial and emissions benefits of AI deployment at sub-500,000 tonne per year scales.
Daniil Matyushenko (Wed,) studied this question.