Additive manufacturing of metals, and in particular Laser Powder Bed Fusion (LPBF), is a promising technology for producing complex geometries. While increasingly established for customized components, its widespread industrial adoption at larger scale is still limited by comparatively high costs, largely caused by low production rates and defects requiring extensive quality control. Improving the LPBF process requires a detailed understanding and predictive modelling capabilities of the physical phenomena occurring at the melt pool scale. Among these, powder denudation plays a key role: vaporization-induced gas flow leads to an inhomogeneous redistribution of powder particles, significantly affecting final part quality. Despite experimental and numerical studies providing valuable insights, a fully resolved computational investigation of gas flow around individual powder particles remains largely unexplored. In this talk, we present a computational approach for predicting powder motion driven by vaporization-induced gas flow in LPBF. The method captures the interplay of multiple physical phenomena, including compressible gas flow, species transport of metal vapor and shielding gas and particle–particle interactions governed by frictional, adhesive contact forces . For the two-way coupled gas–particle dynamics, a discontinuous Galerkin discretization of the compressible Navier-Stokes equations is combined with a Brinkman-type volume penalization method to enforce mechanical interface conditions at particle surfaces, with explicit Runge-Kutta schemes used for time integration. The resulting discrete system is extremely stiff, posing significant numerical challenges such as prohibitively small time-step requirements. We’ll discuss how to cope with these restrictions such that the computation of systems with hundreds of particles remains feasible. Using this approach, we present first simulation results that provide new insights into the mechanisms leading to powder denudation in LPBF. As an outlook, motivated by the coexistence of fast melt-pool flow and slower far-field regimes, we outline regime-tailored coupling strategies to further improve method performance.