In the last two decades much concern has been raised with respect to global dimming and global brightening as a trend in surface solar global irradiance. Surface solar radiation drives various surface processes, like evaporation and photosynthesis, affects surface energy budget and impacts solar energy production technologies and other human activities. In particular, variations of solar resources due to climate change could harm the return of investment for solar energy projects. Therefore, we show a novel method to derive surface direct normal irradiance (DNI) from climate model input data. Since global horizontal (GHI) can be provided simultaneously in a consistent way by our method, this allows to study how projected atmospheric changes may modify GHI and DNI, and hence the output of photovoltaic (PV) and concentrating solar power (CSP) plants. Starting from a given input data set, we then provide concrete projections of solar resources for the middle of the current century. We perform radiative transfer calculations with atmospheric data produced by global climate models. Cloud data is obtained from the model E39C-A, a derivate of the general circulation model ECHAM4, aerosol data from the coupled aerosol-climate model ECHAM5-HAM. For the treatment of clouds in the radiative transfer model, that require particular care since they strongly modulate surface radiation, we introduced the new concept of effective optical thickness and implemented it into E39C-A. The simulations follow the IPCC SRES A1B scenario which describes a future world of rapid economic growth with a balanced use of renewable and fossil fuel power generation. Our calculations indicate that future (2035–2039) surface irradiances are likely to be reduced compared to past (1995–1999) irradiances mainly in southern and western Africa with a decrease of about �20% in DNI and about �5% in GHI. In Europe and Australia solar radiation is likely to increase by about +10% in DNI and some 1–5% in GHI. We observe nearly no change in North America. Climate change has most likely more impact on DNI than on GHI, CSP projects are hence more affected by climate change than PV ventures. The results from the new method can only be as reliable as the input data are. However, the order of magnitude and regional Patterns of our findings are consistent with a recent study that applies other global climate models in combination with a different method to derive surface irradiances. The input data, which our projections strongly rely on, are plausible to the extent that they are qualitatively consistent with IPCC multi-model results for the same scenario as used here.