Despite radio frequency (RF) power amplifiers (PAs) are as old as radio communications, their design is still one of the most difficult electronic circuit design tasks. So, it certainly constitutes one of the fields in which design automation tools will face more difficulties to be launched. Nevertheless, because of the costs associated to the education of a PA design engineer, the many trials and errors required to reach an optimized solution, and its complexity – which hinders systematic design procedures – it is also a field in which design automation would have more impact. So, this work aims to be a step towards this goal. Starting from the optimization of a state-of-the-art PA topology, in which the active device is represented by an artificial neural network, the paper addresses some challenges associated to the desired PA design automation.