Modern cosmology and computing have traveled parallel paths, based on the rapid evolution of semiconductor technology. As sky surveys range wider and deeper across disparate wavelengths, coherently interpreting the measurements they make requires modeling advances that—in many cases—involves the essential use of powerful and balky beasts: supercomputers. As supercomputing steps into and beyond the exascale era, high-fidelity end-to-end modeling of the observed sky is also transitioning from the realms of dreams to reality. Combined with AI-driven approaches, these advances promise rapid acceleration in the pace of obtaining new scientific results. In this talk, I will cover some of the history of these developments and present a number of recent results from work conducted by our group at Argonne National Laboratory.