As global cities accelerate efforts to achieve climate goals, addressing embodied carbon—the emissions embedded in construction materials and processes—has become an urgent yet underexplored challenge. This research proposes a robust, scalable, and data-driven framework to minimize embodied carbon across urban building stocks in the United States. We employed a bottom-up, archetype-based modeling approach combined with a constrained greedy heuristic algorithm to analyze 200,000 buildings in New York and Phoenix, each representing distinct urban typologies. Results show that fewer than 20% of archetypes are responsible for over 80% of total embodied emissions and construction costs, highlighting the effectiveness of targeted retrofit strategies. Our optimization achieved up to 72% reductions in costs and 75% reductions in emissions without altering architectural identities. Moreover, temporal analysis suggests that optimized material flows often correspond with historical infrastructure investment cycles and economic downturns, indicating a potential role for such strategies in bolstering urban economic resilience. We introduce the concept of an “optimization core”—a critical subset of building types whose strategic retrofitting unlocks substantial cost and carbon savings while enhancing flexibility in urban planning.