The objective of the proposed research is to establish a fabrication-to-system co-optimization approach for the development of novel functional oxide thin-film capacitors. Despite the rapid expansion of AI hardware aimed at overcoming the limitations of conventional computing systems, the separation of memory and compute units continues to impose fundamental constraints on scalability and energy efficiency. To address these challenges, this study focuses on exploring scalable M3D computing paradigms using BEOL-compatible functional oxide thin-film capacitors. By establishing a device-to-system benchmarking framework for advanced technology nodes, this work provides a foundation for systematically evaluating emerging devices under future technology nodes. Furthermore, BEOL-integrated memory and power devices that satisfy fabrication constraints and system-level requirements are experimentally demonstrated and engineered at both the device and circuit levels to validate the proposed M3D computing approaches. In particular, a non-volatile capacitive memory device and a high-voltage superlattice capacitor, serving as two important building blocks for M3D computing paradigms, are developed in this work. Through efforts spanning systematic framework development, experimental device demonstration, and device optimization, this research aims to pave the way toward next-generation scalable computing hardware.