HPC infrastructures provide the backbone for advanced numerical simulation, large scale data analytics or machine learning applications. Rapid technology advances, heterogeneity in hardware or power constraints are only a few developments which require HPC hardware to be seamlessly embedded into an environment of extensive user support, training and expert knowledge in various disciplines. In response to these challenges, Germany has recently established the National High-Performance Computing (NHR) Alliance as a long-term nationwide infrastructure, combining operational, training and research expertise of eight universities. The talk first introduces the NHR Alliance and puts it into the context of the German HPC ecosystem. In the second part, recent results from NHR technology exploration activities for the A64FX architecture are presented, including the ECM performance model for A64FX and a discussion of performance characteristics of two important compute kernels (sparse matrix-vector multiplication and a domain wall kernel from quantum chromodynamics)