Distinct grain structures have been observed in Metal Additive Manufacturing (MAM) processes. These grain structures feature columnar grains which occasionally mix with equiaxed grains. The occurrence of these grain structures is not yet fully understood. In this work, direct laser deposition process is studied as a typical MAM process. A finite volume model is first implemented to obtain the thermal history. Next, the thermal history is fed into a Cellular Automata (CA) model to simulate the epitaxial and competitive growth through which the columnar grains are formed. Nucleation is included in the model to predict the generation of equiaxed grains, and is characterized by two nucleation parameters, the nucleation density and the critical undercooling. The simulation results show that both the nucleation parameters and process parameters can significantly affect the grain structure. The simulated grain structures examined on different planes can be significantly different, revealing the complexity of the 3-dimensional grain structures in MAM processes.