Localized surface plasmon resonances (LSPRs) associated with metallic nanostructures offer unique possibilities for light concentration beyond the diffraction limit, which can lead to strong field confinement and enhancement in deep subwavelength regions. In recent years, many transformative plasmonic applications have emerged, taking advantage of the spectral and spatial tunability of LSPRs enabled by near-field coupling between constituent metallic nanostructures in a variety of plasmonic metastructures (dimers, metamolecules, metasurfaces, metamaterials, etc.). For example, the “hot spot” formed at the interstitial site (gap) between two coupled metallic nanostructures in a plasmonic dimer can be spectrally tuned via the gap size. Capitalizing on these capabilities, there have been significant advances in plasmon enhanced or enabled applications in light-based science and technology, including ultrahigh-sensitivity spectroscopies, light energy harvesting, photocatalysis, biomedical imaging and theranostics, optical sensing, nonlinear optics, ultrahigh-density data storage, as well as plasmonic metamaterials and metasurfaces exhibiting unusual linear and nonlinear optical properties. In this review, we present two complementary approaches for fabricating plasmonic metastructures. We discuss how meta-atoms can be assembled into unique plasmonic metastructures using a variety of nanomanipulation methods based on single- or multiple-probes in an atomic force microscope (AFM) or a scanning electron microscope (SEM), optical tweezers, and focused electron-beam nanomanipulation. We also provide a few examples of nanoparticle metamolecules with designed properties realized in such well-controlled plasmonic metastructures. For the spatial controllability on the mesoscopic and macroscopic scales, we show that controlled self-assembly is the method of choice to realize scalable two-dimensional, and three-dimensional plasmonic metastructures. In the section of applications, we discuss some key examples of plasmonic applications based on individual hot spots or ensembles of hot spots with high uniformity and improved controllability.