DescriptionIn this talk, we will discuss different aspects of the path towards exascale from the perspective of the Juelich Supercomputing Centre (JSC) in Germany. The purpose for targeting next levels of performance is the enablement of new science. We therefore understand exascale as the ability to provide 2 orders more performance for scientific applications compared to today. This does not necessarily translates into a corresponding scale-up of the throughput of floating-point operations. We will look into the application domains that play a key role for JSC in future, analyze selected requirements and confront this with current technology roadmaps.