Selective Laser Sintering (SLS) and Selective Laser Melting (SLM) are parent Additive Manufacturing processes using a laser to solidify metallic, ceramic, polymer or composite powders. During the process, the object is built layer by layer. A laser source is responsible for the consolidation, by local heating. The light is deviated by a scanning head according to the instructions of an STL file. Then partial fusion of the particles takes place, followed by a solidification of the liquid created. Kinetics of these steps are very high and play an important role in the final microstructure (rearrangement of the particles, pore creation, residual stresses). In the case of polymers and composites the technique is now well understood and widely used, but for metals and ceramics it suffers from a lack of precision, surface roughness and poor mechanical properties. The goal of this Ph.D. work is to understand the effect of the thermal gradients on the consolidation process, using different laser parameters (power, pulse frequency, scan speed,…). In SLS, microscopic thermal gradients arise, due to the fact that the laser is pulsed and that only the exterior of the powder particles is molten. A thermal model to describe the interaction between a laser beam and a (spherical) grain is proposed. This model allows for the incorporation of the latent heat of fusion and for a realistic surrounding. The absorbed laser intensity is modeled by means of the Mie theory. Experiments where two particles of powder are isolated and illuminated by the laser are carried out in order to measure the interparticular necks and the volume of liquid formed for different repetition rates. The thermal model leads to good predictions of the particles final sintering state. Fluid flow models are investigated in order to determine the dynamic of the molten liquid. The main issue is to explain the capillary flow mechanisms leading from the molten material to the neck formation. Two models are derived to simulate the fluid flow between the particles. The first one simulates a capillary flow between two parallel plates. The second one deals with energetic considerations arising from Frenkel's principle. In both cases, the final output is the liquid life-time necessary to get the interparticular neck lengths experimentally observed. We point out that the two models predict the same liquid lifetime, although they are derived with different hypotheses. In SLM, continuous lasers are used and one has to deal with macroscopic thermal gradients, since the particles are completely molten. A very high laser power is used and the negative thermal effects (like thermal stresses or balling effects) are important. A solution to control them is to adjust the scanning strategy. Four scanning strategies are investigated for material with different thermal conductivities. The three-dimensional model used to describe the laser-matter interactions and the temperature evolution of the scanned powder bed allows for a finite latent heat (Stefan-problem) and for conductivity modifications due to the consolidation. We show that this finite element thermal model can be efficiently used to anticipate most of the problems (like cracks or balling) arising in practice. The benefit of avoiding thermal gradients is shown, in particular by EBSD analysis. Finally, applications of the SLS/SLM technique on different pieces built during this work are shown. A new way to build support structure is also proposed.