State-of-the-art algorithms for energy-efficient power allocation in wireless networks are based on fractional programming theory, and allow to find the global maximum of the energy efficiency only in noise-limited scenarios. In interference-limited scenarios, several sub-optimal solutions have been proposed, but an efficient framework to globally maximize energy-efficient metrics is lacking. The goal of this work is to fill this gap by making use of fractional programming theory jointly with monotonic optimization. The resulting optimization framework is useful for at least two main reasons. First, it sheds light on the ultimate energy-efficiency performance of wireless networks. Second, it provides the means to benchmark the energy efficiency of state-of-the-art, but sub-optimal, solutions.