As CMOS scaling becomes more and more challenging there is strong impetus for beyond CMOS device research to add new functionality to ICs. In this article, a promising technology with non-volatile ferro-magnetic computing states - the so-called Perpendicular Nanomagnetic Logic (pNML) - is reviewed. After introducing the 2D planar implementation of NML with magnetization perpendicular to the surface, the path to monolithically 3D integrated systems is discussed. Instead of CMOS substitution, additional functionality is added by a co-processor architecture as a prospective back-end-of-line (BEOL) process, where the computing elements are clocked by a soft-magnetic on-chip inductor. The unconventional computation in the ferromagnetic domain can lead to highly dense computing structures without leakage currents, attojoule dissipation per bit operation and data-throughputs comparable to state-of-the-art high-performance CMOS CPUs. In appropriate applications and with specialized computing architectures they might even circumvent the bottleneck of time-consuming memory access, as computation is inherently performed with non-volatile computing states. (C) 2015 Elsevier Ltd. All rights reserved.