The increasing complexity of modern power systems, driven by large-scale integration of renewable energy sources, electrification and decentralized resources, is challenging traditional approaches to system operation and planning. Digital Twins (DTs) have emerged as a promising paradigm to support data-driven decision-making by linking physical systems with digital models. However, realizing DTs in power systems remains challenging due to the need to integrate heterogeneous models, operate across multiple spatial and temporal scales, and ensure interoperability between diverse stakeholders and tools. This keynote discusses the role of DTs in future power grids, building on recent advances in power system modelling and co-simulation. It highlights key challenges related to data integration, model coupling and the joint representation of physical grid dynamics and market mechanisms. Particular attention is given to the transition from isolated models towards more interoperable DT setups, requiring appropriate architectural choices, standardized interfaces and data exchange approaches. The talk provides a system-level perspective on DT engineering and outlines directions for developing more scalable and interoperable DT frameworks, with the aim of supporting improved analysis and coordination in future energy systems.