Cast Duplex Stainless Steels (CDSS) have been extensively used as structural components in light water reactor (LWR) power plants as primary coolant pipes and pump casings. These large components are in persistently extreme and damaging environments, including high temperature, high pressure, corrosive exposure, and low dose radiation for extended periods of time. This work focuses on the thermal embrittlement of CDSS that leads to poor mechanical properties and is a concern in the extended-term operation of nuclear power plants. Many past studies have concluded that the miscibility gap in the Fe-Cr phase diagram leads to spinodal decomposition, which is the primary source for thermal embrittlement of the δ-ferrite phase in CDSS. The hardening caused by spinodal decomposition hinders plastic deformation and thus leads to brittle fracture. In this work, mechanical testing and microstructural characterization, mainly by Atom Probe Tomography (APT), were performed on short-term (up to 2 years) thermally aged CDSS at accelerated temperatures to understand the embrittlement process and mechanisms of thermal-aging degradation. Two CDSS materials, CF–3 and CF–8, were thermally aged for 4300 hours (h), 8600 h, 12900 h, and 17200 h at 280 oC, 320 oC, 360 oC, and 400 oC. Both alloys exhibited an increase in hardness and associated reduction in impact energy with increased aging time across all temperatures. The unparalleled insight of the APT technique provided a high resolution analysis of 3-D distributions of atoms to experimentally characterize the morphology of phase separation. It allowed for the analysis of M23C6 carbides and nanostructured clusters, including the spinodal decomposition of the δ-ferrite into networked α/α' domains and precipitation of G–phase at the α/α' domain interface. These transformations are shown to be highly dependent on the initial composition and aging temperature. From the wavelength (size) and amplitude (composition fluctuation) measurements, we successfully calculated the activation energies and inferred the progression of dominating diffusion mechanism during spinodal decomposition and G–phase precipitation. This data may potentially be used to predict the reliability of LWR components after many years of service.