The anticipated increase in the frequency of extreme weather events and the growing photovoltaic (PV) penetration in the energy system raise concerns about future variability in PV power generation. However, future changes in global PV variability based on the CMIP6 still remain uncertain. To fill this gap, data from 16 state-of-art CMIP6 models were employed to analyze trends in both PV power generation and variability for 2025–2100 under low, medium, and high greenhouse gas emission scenarios. A multivariate bias correction algorithm (MBCn) was used to calibrate CMIP6 data, with reanalysis data - ERA5 serving as the benchmark. Additionally, the impact of meteorological factors were examined. The results indicate that under the low emission scenario, global PV power generation shows a slight increase, while the other two scenarios exhibit decreasing trends. In terms of PV variability, changes are correlated with latitude, with high-latitude regions more likely to face higher fluctuations, leading to an additional approximately 16% of low-power days by 2100 under the high emission scenario. Furthermore, over two-thirds of the land area experiences a decrease in PV power generation along with increased variability under the high emission scenario, approximately 8 times higher than the low emission scenario. Considering the impact of meteorological factors, removing days with extreme high and low solar irradiance increases stability by about 23%. However, eliminating days with extreme high and low temperature worsens stability, revealing that global warming reduces variability to some extent due to the opposing effects of temperature and irradiance on PV power generation. This study highlights the consequences of climate change on PV power generation variability, providing valuable insights for PV installation planning, especially for countries at higher latitudes.