Silicon has been the semiconductor of choice for microelectronics largely because of the unique properties of its native oxide (SiO2) and the Si/SiO2 interface. For high-temperature and/or high-power applications, however, one needs a semiconductor with a wider energy gap and higher thermal conductivity. Silicon carbide has the right properties and the same native oxide as Si. However, in the late 1990's it was found that the SiC/SiO2 interface had high interface trap densities, resulting in poor electron mobilities. Annealing in hydrogen, which is key to the quality of Si/SiO2 interfaces, proved ineffective. This paper presents a synthesis of theoretical and experimental work by the authors in the last six years and parallel work in the literature. High-quality SiC/SiO2 interfaces were achieved by annealing in NO gas and monatomic H. The key elements that lead to highquality Si/SiO2 interfaces and low-quality SiC/SiO2 interfaces are identified and the role of N and H treatments is described. More specifically, optimal Si and SiC surfaces for oxidation are identified and the atomic-scale processes of oxidation and resulting interface defects are described. In the case of SiC, we conclude that excess carbon at the SiC/SiO2 interface leads to a bonded Si-C-O interlayer with a mix of fourfold- and threefold-coordinated C and Si atoms. The threefold coordinated atoms are responsible for the high interface trap density and can be eliminated either by H-passivation or replacement by N. Residual Si-Si bonds, which are partially passivated by H and N remain the main limitation. Perspectives for the future for both Si- and SiC-based MOSFETs are discussed.