A new processing architecture for semiconductor nanowire grid fabrics is presented. The system consists of a large number of functionally identical units called cells. Cells are locally interconnected with nearest neighbors, with a limited number of global signals routed from supporting CMOS circuitry. One possible implementation of a digital Cellular Neural Network (CNN) using this architecture is shown. The digital cellular design may be up to 27X denser than an equivalent 18nm CMOS implementation. The system, based on a collective computation model, may also significantly alleviate manufacturing, since 100% fault-free components may not be necessary.