A prescriptive optimization framework for designing sustainable lithium supply portfolio

Lithium is central to energy transition as the key element of electric vehicles and grid storage, but its expanding is accompanying with worsen environmental pressures. Existing studies fall short in exploring the full mitigation potential of environmental impacts due to the reliance on a descriptive approach of comparing predefined options. Here we develop an artificial intelligence-driven life cycle assessment methodology to assess 18-dimensional environmental performance for global 121 mining sites, and further optimize global lithium supply portfolios from 2025 to 2050 across four demand scenarios. The optimization model enforces real-world constraints for project commissioning, capacity ramp-up and resource depletion to enable actionable insights. Results indicated that optimized portfolios lower cumulative burdens across 18 environmental dimensions by 53.3–61.8% on average compared to the baseline scenario, with the most reliable gains for climate and health impacts. While water use falls only modestly, and mineral resource scarcity can worsen as other impacts decline. Optimized portfolios suggest a brine-based supply by Chile’s Atacama and Argentina’s Cauchari-Olaroz, Olaroz, and Tres Quebradas, followed around 2035 by accelerated Australian spodumene (Pilgangoora, Mt Holland, Greenbushes) and diversification from other countries. The results point to practical levers for industry and policy stakeholders to prioritize supply sources to align surging lithium demand with multi-dimensional environmental goals.