Traditionally, computing systems have relied on scaling of transistor feature sizes for enhancing energy efficiency, throughput, performance, functional density, and most importantly cost (per component). Unfortunately the benefits gained from further scaling are greatly reduced. Future computing systems need to overcome the fundamental efficiency‐robustness barrier in order to continue to have societal‐scale impact. A key requirement is to envision this future as a convergence of three domains – emerging applications, alternative computational models and architectures, and beyond CMOS nanofabrics. A compelling vision of future computing systems would be one where the application level metrics are accounted for during design, where statistical, hysteretic, and other attributes of nanoscale fabrics could be exploited for designing computational primitives such as nanofunctions required by these applications, and where alternative models for computing could be leveraged to design systems meeting application‐level requirements. Achieving this vision calls for a journey from systems‐to‐nanofabrics and back.