Heterogeneous catalysts, often consisting of metal nanoparticles supported on high-surface-area oxide solids, are common in industrial chemical reactions. Researchers have increasingly recognized the importance of oxides in heterogeneous catalysts: that they are not just a support to help the dispersion of supported metal nanoparticles, but rather interact with supported metal nanoparticles and affect the catalysis. The critical role of oxides in catalytic reactions can become very prominent when oxides cover metal surfaces forming the inverse catalysts. The source of the catalytic activity in homogeneous catalysts and metalloenzymes is often coordinatively unsaturated (CUS) transition metal (TM) cations, which can undergo facile electron transfer and promote catalytic reactions. Organic ligands and proteins confine these CUS cations, making them highly active and stable. In heterogeneous catalysis, however, confining these highly active CUS centers on an inorganic solid so that they are robust enough to endure the reaction environment while staying flexible enough to perform their catalysis remains a challenge. In this Account, we describe a strategy to confine the active CUS centers on the solid surface at the interface between a TM oxide (TMO) and a noble metal (NM). Among metals, NMs have high electron negativity and low oxygen affinity. This means that TM cations of the oxide bind strongly to NM atoms at the interface, forming oxygen-terminated-bilayer TMO nanostructures. The resulting CUS sites at the edges of the TMO nanostructure are highly active for catalytic oxidation reactions. Meanwhile, the strong interactions between TMOs and NMs prevent further oxidation of the bilayer TMO phases, which would otherwise result in the saturation of oxygen coordination and the deactivation of the CUS cations. We report that we can also tune the oxide-metal interactions to modulate the bonding of reactants with CUS centers, optimizing their catalytic performance. We review our recent progress on oxide-on-metal inverse catalysts, mainly the TMO-on-Pt (TM = Fe, Co, and Ni) systems and discuss the interface-confinement effect, an important factor in the behavior of these catalytic systems. We have studied both model catalyst systems and real supported nanocatalysts. Surface science studies and density functional theory calculations in model systems illustrate the importance of the oxide-metal interfaces in the creation and stabilization of surface active centers, and reveal the reaction mechanism at these active sites. In real catalysts, we describe facile preparation processes for fabricating the oxide-on-metal nanostructures. We have demonstrated excellent performance of the inverse catalysts in oxidation reactions such as CO oxidation. We believe that the interface confinement effect can be employed to design highly efficient novel catalysts and that the inverse oxide-on-metal catalysts may find wide applications in heterogeneous catalysis.