The impact of climate change on power demand in Japan is a matter of concern for the Japanese authorities and power companies as it may have consequences on the power grid. We trained random forest models against daily power data in ten Japanese regions and for different types of power generation to project changes in future power production and its carbon intensity. To do so, we used twelve predictors: six climate variables, five variables accounting for human exposure to climate, and one variable for the level of human activities. We then used the models trained from the present-day period to estimate the future power demand, carbon intensity, and pertaining CO2 emissions over the period 2020-2100 under three SSPs scenarios (Shared Socioeconomic Pathways: SSP126, SSP370, and SSP585). The impact of climate change on CO2 emissions via power generation shows seasonal and regional disparities. In cold regions, a decrease in power demand during winter under future warming leads to an overall decrease in power demand over the year. In contrast, the decrease in winter power demand in hot regions can be overcompensated by an increase in summer power demand because of more frequent hot days, leading to an overall annual increase. From our regional models, the power demand should increase the most in most Japanese regions in May, June, September, and October and not in the middle of summer, as has been found in older studies. Such an increase could result in regular power outages during those months if not considered, as the power grid could be particularly tense. Overall, we observed that power demand in regions with extreme climates is more sensitive to global warming than in temperate regions. The impact of climate change on power demand induces a net annual decrease in CO2 emissions in all regions except for Okinawa, in which power demand strongly increases during the summer, resulting in a net annual increase in CO2 emissions. However, climate change's impact on carbon intensity may reverse the trend in some regions (Shikoku, Tohoku). We also assessed the relative impacts of socioeconomic factors such as population, GDP, and environmental policies on CO2 emissions. When combined with these factors, we found that the climate change effect is more important than when considered individually and significantly impacts total CO2 emissions under SSP585.