This paper presents an efficient sub-stepping strategy for part-scale thermo-mechanical simulations of Metal Additive Manufacturing (MAM) processes, with a particular focus on Laser Powder Bed Fusion (LPBF). The proposed approach aims to accelerate high-fidelity thermo-mechanical analyses by partitioning the computational domain into a localized region surrounding the thermo-mechanically affected zone (TMAZ), which evolves dynamically according to the actual scanning sequence, and a global region representing the remaining part [1]. The local (fast) and global (slow) regions are advanced in time using different time steps through a Robin–Robin sub-stepping scheme, enabling an effective treatment of the strong time-scale disparity inherent to LPBF processes. This local–global strategy significantly reduces computational cost while preserving the accuracy of the thermal and mechanical predictions [2]. Comparisons with full-order finite element simulations demonstrate that the proposed method achieves substantial efficiency gains without compromising solution fidelity. Overall, the framework enables accurate and scalable simulation of MAM processes, supporting process qualification and industrial deployment in sectors such as aerospace, automotive, and energy.