Molecular dynamics simulations are capable of resolving molecular recognition processes with chemical accuracy, but their practical application is popularly considered limited to the timescale accessible to a single simulation, which is far below biological timescales. In this perspective article, we propose that the true limiting factor for molecular dynamics is rather the high hardware and electrical power costs, which constrain not only the length of runs but also the number that can be performed concurrently. As a result of innovation in accelerator processors and high-throughput protocols, the cost of molecular dynamics sampling has been dramatically reduced and we argue that molecular dynamics simulation is now placed to become a key technology for in silico drug discovery in terms of binding pathways, poses, kinetics and affinities.