This article gives an overview of the current state-of-the-art of photovoltaic electricity technology, and addresses its potential for cost reductions over the first few decades of the 21st century. Current PV production cost ranges are presented, both in terms of capacity installation and electricity generation, of single crystalline silicon, multi-crystalline silicon, amorphous silicon and other thin film technologies. Possible decreases of these costs are assessed, as expected according to the learning-curve methodology. We also estimate how much PV could gain if external costs (due to environmental and health damage) of energy were internalised, for example by an energy tax. Our conclusions are that, (1) mainly due its high costs, PV electricity is unlikely to play a major role in global energy supply and carbon emissions abatement before 2020, (2) extrapolating learning curves observed in the past, one can expect its costs to decrease significantly over the coming years, so that a considerable PV electricity share world-wide could materialise after 2020, (3) niche-market applications, e.g. using stand-alone systems in remote areas, are crucial for continuing 'the tide along the learning curve', (4) damage costs of conventional (fossil) power sources are considerable, and their internalisation would improve the competitiveness of PV, although probably not enough to close the current cost gap. (C) 2003 Elsevier Science Ltd. All rights reserved.