In fractal compression a signal is encoded by the parameters of a contractive transformation whose fixed point (attractor) is an approximation of the original data. Thus fractal coding can be viewed as the optimization problem of finding in a set of admissible contractive transformations the transformation whose attractor is closest to a given signal. The standard fractal coding scheme based on the collage theorem produces only a suboptimal solution. We demonstrate by a reduction from MAXCUT that the problem of determining the optimal fractal code is NP-hard. To our knowledge, this is the first analysis of the intrinsic complexity of fractal coding. Additionally, we show that standard fractal coding is not an approximating algorithm for this problem.