must be considered. This branch of research involves the so-called Moore than More's devices. [1] Some attempts to ease consumption rely on on-chip integration of nanoscale energy sources, [2,3] while others, notably in memory computing (IMU), [4] offer a revisited architecture where computation is decentralized. Other research groups proposed the integration of energy sources in close proximity of the memory, [5,6] or completely integrated within the memory architecture. [7] In this work, we explore a radically different concept, which lies on "dualbehavior" devices, able to store either information or energy, depending on the applied bias. Such capability would be greatly beneficial, allowing localized and high bandwidth energy supply to the processing unit (the memory or a dedicated arithmetic logic unit, ALU). The samples considered in our study indeed resemble ionic batteries at the nanoscale, providing ground to our inquiry of using these devices as energy sources, other than memory cells. Their operation relies on faradaic processes; therefore, the resulting energy density is expected to well exceed that of electrostatic capacitors, possibly being comparable to supercapacitors'. [8] The diameter of the devices under study can range between 1 and 100 µm, resulting in much smaller than the diameter of state-of-the-art planar supercapacitors (≈mm 2 ), [2] making such architecture more scalable and granular than any other state-of-the-art integrated power source. Energy storage is achievable when the device is, under a memory point of view, storing a logic 0, and could be accumulated during low logic operation activity for later use, for example, during the most power-hungry phases. These devices would also offer the advantage of placing the battery cell in close proximity to the target, meaning reduced IR drop and voltage undershoot, which develop in a typical inductive-impedance power delivery network (PDN). [9] A broad range of applications could be envisioned, each demanding different energy requirements, with widespread specifications. The most suitable target field should be selected taking into account the output voltage, energy, and power delivered by such RRAM-based batteries. In this work, we provide an outlook on some possible implementations, with some quantified ranges in terms of energy and instantaneous power requirements. Figure 1 reports the three main eligible domains: energy to memory, [10] energy to logic, [11] and neuromorphic computing. [12][13][14][15] This work explores the innovative concept of a hybrid dual-behavior device, based on emerging nonvolatile memory technology, for both data retention and energy storage. RRAM (resistive random access memory) is considered a major candidate as next-generation memory, thanks to its promising performances in terms of scalability and CMOS process compatibility. Its working mechanisms, based on faradaic processes, motivate the study on the feasibility of operating RRAM also as energy storage element. To evaluate the energy capability, various el...