We investigate the interpolation of power spectra of matter fluctuations using artificial neural networks (PkANN). We present a new approach to confront small-scale non-linearities in the power spectrum of matter fluctuations. This ever-present and pernicious uncertainty is often the Achilles heel in cosmological studies and must be reduced if we are to see the advent of precision cosmology in the late-time Universe. We show that an optimally trained artificial neural network (ANN), when presented with a set of cosmological parameters ( and redshift z), can provide a worst-case error ≤1 per cent (for z≤ 2) fit to the non-linear matter power spectrum deduced through N-body simulations, for modes up to k≤ 0.7 h Mpc−1. Our power spectrum interpolator is accurate over the entire parameter space. This is a significant improvement over some of the current matter power spectrum calculators. In this paper, we detail how an accurate interpolation of the matter power spectrum is achievable with only a sparsely sampled grid of cosmological parameters. Unlike large-scale N-body simulations which are computationally expensive and/or infeasible, a well-trained ANN can be an extremely quick and reliable tool in interpreting cosmological observations and parameter estimation. This paper is the first in a series. In this method paper, we generate the non-linear matter power spectra using halofit and use them as mock observations to train the ANN. This work sets the foundation for Paper II, where a suite of N-body simulations will be used to compute the non-linear matter power spectra at sub-per cent accuracy, in the quasi-non-linear regime (0.1 ≤k≤ 0.9 h Mpc−1). A trained ANN based on this N-body suite will be released for the scientific community.