In this paper, we analyze the technical efficiency and CO2 reduction potentials of German power and heat plants, using a non-parametric sequential Data Envelopment Analysis. We apply a metafrontier framework to evaluate plant-level efficiency in the transformation of inputs into desirable (energy) and undesirable (CO2 emissions) outputs, taking into account different fossil fuel generation technologies. We use a unique data set of coal-, lignite-, gas- and biomass-fired power plants from 2003 through 2010 that provides an unbalanced panel of 1459 observations; the results are also checked against a balanced panel with a smaller number of observations. Although we find intra-group differences within energy generation technology, natural gas fired power plants clearly have the highest efficiency. Furthermore, the analysis points to significant savings potentials for CO2 and fuel-input, and derives policy conclusions for the ongoing electricity sector reformation.