The Spinning Enhanced Visible and Infrared Imager (SEVIRI) measurements from the Meteosat Second Generation (MSG) satellites enable global monitoring of the distribution of clouds during day and night, with a spatial, temporal and spectral resolution that allows for better understanding of the role of clouds in global radiation budget and in climate in general. A method to retrieve cloud properties from nighttime SEVIRI measurements is described in this paper. The method is applicable to single-layer water clouds over sea surfaces and it is based on the inversion of a forward theoretical radiative transfer model, that simulates the radiances reaching the SEVIRI infrared detectors from a specified configuration of the earth–cloud– atmosphere system. This model accounts for scattering and absorption processes in the assumed horizontally homogeneous adiabatic cloud layer. For the inversion of this model, artificial neural networks techniques have been used in this work. The main advantage that these techniques provide is their low computational cost, which makes them suitable for the implementation of operational retrieval procedures. Results obtained by the proposed method are compared with the values provided by the CloudSat derived 2B-TAU product, and those derived from NOAA-AVHRR nighttime imagery, obtaining good agreements.