Metal-based additive manufacturing (AM) is revolutionizing the production process and introducing unprecedented capabilities, which are quickly becoming indispensable across a wide range of industries. Direct Energy Deposition (DED) in particular exhibits a high potential for space applications due to no imposed limitation on the size of manufactured objects and the ability to operate in micro-gravity conditions. DED however remains hindered by poor deposition quality and reproducibility, which appear to originate in the powder stream condition. Increased accuracy of the blown powder dynamics hence represents a crucial ingredient of next-generation DED models. Powder stream is usually modelled with the use of computational fluid dynamics (CFD) as a two-phase flow problem involving a dispersed second phase [5]. Powder particle collisions and their interaction with the melt-pool cannot be accounted for by these models and are regularly disregarded on account of these particles occupying a small volume fraction in the carrier gas flow. This assumption was put to the test using a Discrete Element Method (DEM) model of the particle stream of a discrete coaxial nozzle. While neglecting the interaction between carrier gas and powder particles, the results showed that non-negligible portions of powder grains are involved in grain collisions with substantial rebound angles, which underlined the need to account for inter-particle interaction in DED stream models. This sparked the development of a fully coupled CFD-DEM model of powder stream in DED, using state-of-the-art approaches for parallelly resolving the fluid and the discrete phase dynamics, accounting for the drag on the powder particles as well as the resulting reactive force on the fluid. The model is thus capable of accounting for the full set of granular dynamics, including grain collisions as an essential influence on powder stream dynamics.