As the space industry is undergoing an evolution under the paradigms of “New Space Economy” and the “Future Space Ecosystem”, it has become increasingly evident that the current industry-standard approaches toward design, development, and qualification of the Electric Propulsion (EP) systems, which are largely based on an empirical “trial-and-error” methodology, cannot address the emerging needs and keep abreast of the rapid changes in the market trends. Furthermore, as Artificial Intelligence (AI) is proliferating through the space industry toward the realization of the next-generation autonomous near-Earth satellites and interplanetary spacecrafts, the conventional monitoring and control strategies of the EP systems become inadequate and need to give way to approaches that are aligned with the satellite-level autonomy requirements. A digital twin – a technology capable of providing an accurate virtual representation of a physical asset – is a game-changing concept that catalyzes the transcendence of the EP industry past the pressing challenges it is facing today. In this paper, we first aim to: (i) define digital twins with a systematic and scientific mindset, highlighting in particular how they surpass traditional modelling, (ii) enumerate digital twins’ breakthrough promises for the EP and the broader space industry, and (iii) specify the challenges to realize fully fledged, practical, and scalable digital twins for EP. Following these, we briefly report on the technical progress achieved and/or planned at Imperial Plasma Propulsion Laboratory to fill the foundational gaps in the three building block elements of digital twins, namely, (i) a cost-effective kinetic particle-in-cell (PIC) model to generate extensive high-fidelity databases for machine learning (ML), (ii) ML-enabled, fast reduced-order models (ROMs) for performance predictions, and (iii) a digital twin architecture that supports the integration of numerical models and provides data pipelines and interfaces for the digital twin’s data exchanges with the real world and its dynamic updating.