Among the materials available today for additive manufacturing (AM), high-performance thermoplastics reinforced with short carbon fibers allow the unique possibility of controlling the orientation of these reinforcements inside a part. This makes 3D printed carbon fiber reinforced polymers (CFRP) a serious contender for replacing heavier metallic parts. But despite its high potential, AM represents a significant paradigm shift in the field of mechanical design and several challenges must be addressed before it sees widespread adoption for industrial-scale applications. In the case of CFRP, the heterogeneity of the solid and its resulting anisotropy have been identified as crucial challenges for modelling. The absence of software packages that can accurately account for the peculiarities of this material is a problem for designers, who are forced to use overly conservative safety factors to ensure that the end result satisfies load requirements. This project aims at implementing a customized version of Moulinec and Suquet’s Fast Fourier Transform (FFT) formulation of the Lippmann-Schwinger equation, in which tomographic scans of real material are the starting point to an iterative resolution of the elasticity equations. This method is made computationally efficient by relying heavily on the FFT and can surpass the finite element method (FEM) by evacuating the need for meshing. First a benchmarking of the existing software packages that implement the FFT algorithm is conducted with numerically-generated microstructures, to map the performance and the validity of the outputs as compared to an FEM package (Abaqus). Simulations are performed for configurations that are known to be hard to model, i.e: high volume fraction of inclusion, high aspect ratio, high properties contrast, and in the presence of porosity. Then, tomographic scans of real 3D-printed CFRPs are used, and the predictions of both FFT and FEM methods are compared to the measured properties of CFRPs.