Among the techniques widely used in CEM (Computational Electromagnetics), the FDTD method (Finite-Difference Time Domain) is one of the most demanding in computational resources. Despite the computing power available today on a personal computer, industry demands the analysis of large and complex models to be used in EMC (ElectroMagnetic Compatibility) applications among others. In the last decade, the main strategies used to improve the performance of FDTD have been (1) parallelizing FDTD in clusters of computers, (2) designing digital systems which act as accelerators using reconfigurable hardware, and (3) incorporating Graphics Processing Units (GPU) to accelerate computations. This paper presents a brief review of the state of the art in HPC (High Performance Computing) in FDTD, foresees the trend in HPC for CEM and explains the behaviour of algorithm in one of these architectures.