Ice-shelf thickness is an important boundary condition for ice-sheet and sub-ice-shelf cavity modelling. It is required near the grounding line to calculate the ice flux used to determine ice-sheet mass balance by comparison with the upstream accumulation. In this mass budget approach, the accuracy of the ice thickness is one of the limiting factors in the calculation. We present a satellite retrieval of the ice thickness for all Antarctic ice shelves using satellite radar altimeter data from the geodetic phases of the European Remote-sensing Satellite (ERS-1) during 1994–95 supplemented by ICESat data for regions south of the ERS-1 latitudinal limit. Surface elevations derived from these instruments are interpolated on to regular grids using kriging, and converted to ice thicknesses using a modelled firn-density correction. The availability of a new spatial variable firn-density correction significantly reduces the error in ice thickness as this was previously the dominant error source. Comparison to airborne data shows good agreement, particularly when compared to SOAR CASERTZ data on the largest ice shelves. Biases range from −13.0 m for areas where the assumption of hydrostatic equilibrium breaks down, to 53.4 m in regions where marine ice may be present.