A numerical two scale method for the prediction of tensile and bending stiffness and strength of medium density fiberboards (MDF) is proposed with the aim to study the fiber orientation influence on mechanical properties of MDF. The method requires less experi- mental data to optimize MDF and to improve industrial manufacturing technology of MDF. A new method for computing orientation tensors of the compressed fiber network is proposed. First, the virtual microstructure is generated by simulations of a fiber lay- down and a subsequent compression to obtain the necessary density. The density profile, fiber length, thickness, and orientation are used for the microstructure generation, which are obtained from lCT images and image analysis tools. Then a new damage model for the wood fiber cell walls and joints is introduced. The microstructural problem is formu- lated as a Lippmann–Schwinger type equation in elasticity and solved by using Fast Fourier Transformation (FFT). The macroscopic three point bending test is simulated with hexahe- dral finite elements and analytical methods based on Euler–Bernoulli theory. The differ- ence between bending strength and stiffness numerically obtained and corresponding experimentally measured values is less than 10%. This study lays a foundation for the opti- mal design of MDF fiber structures and the optimization of industrial manufacturing pro- cesses. The first results show an increase of up to 60% for bending stiffness in the case of strongly oriented fibers.