Driven by standardization and commercialization, digital infrastructures evolve in waves. Over the last few years, a particular focus has been on realizing ultra-reliable low-latency wireless communications (URLLC), anticipated mostly for rather specific use cases in industrial automation. Even though initial such systems finally exist today - with future network releases advancing URLLC capabilities even more - the broad market impact to date is low. We argue in this paper that an essential missing component for corresponding dependable applications like closed-loop control or human-in-the-loop are nearby compute capabilities provided within the infrastructure, aka edge computing capabilities. Only in conjunction can such future infrastructures support dependable applications to a full extent. Nevertheless, this also leads to unique challenges which will be central to the evolution of networked infrastructures during the current decade. Out of this evolution of networked infrastructures, we finally argue that a new type of networked application class will emerge, resembling the representation of various aspects of reality in the infrastructure at any point in time. We dub this development the Internet of Reality, and discuss further challenges in this context.