The recently demonstrated ability to perform accurate, pre- cise and rapid computation of ligand binding strengths has driven interest in the use of molecular simulations for applications in both computer- aided drug design and patient specific medicine. Simulation protocols based on ensembles of multiple runs of the same system provide an effi- cient method for producing robust free energy estimate, and equally im- portant statistical uncertainties. Variations in the chemical and biophys- ical properties of different systems impact the optimal protocol choice for different proteins and classes of drugs targetting them. However, these are difficult to determine a priori, thus requiring adaptive execution of a simulation workflows. We introduce the High-throughput Binding Affin- ity Calculator (HTBAC) to enable the scalable, adaptive and automated calculation of the binding free energy on high-performance computing re- sources. We investigate weak scaling behaviour for screening sixteen drug candidates concurrenlty using thousands of multi-stage pipelines on more than 32,000 cores. This permits a rapid time-to-solution that is essen- tially invariant with respect to the calculation protocol, size of candidate ligands and number of ensemble simulations. As such, HTBAC advances the state of the scale and sophistication of binding affinity calculation. HTBAC uses readily available building blocks to attain both workflow flexibility and performance; our scaling experiments are performed on the Blue Waters machine at NCSA.