In this talk, I will start with perspectives on the unique multifunctional feature of spintronic devices as an enabler for many new applications from MESO, CoMET, stochastic and probabilistic computing devices to computational random access memory (CRAM). Then I will focus on one specific discussion of our experimental demonstration and first-principles calculation of quantum confinement in sputtered topological spin-orbit torque (SOT) materials. Electronic band-structure analysis indicates that the reduced dimensionality and quantum confinement strongly influences the spintronic properties of the TI. Our theory identifies the presence of lowly dispersive surface bands with large charge-to-spin conversion efficiency in nanoscale grains, which explain the experimentally observed enhancement in the figure-of-merit. I will further report here the modeling and benchmarking for SOT-RAM based CRAM vs other near and in memory computing for different applications including pattern recognition, binary-neural-net-based image classification, and FFT.