The redshift evolution of the X-ray emission of galaxies is an excellent diagnostic for the cosmological evolution of compact object populations and testing models for binary stellar evolution. These in turn are key ingredients for understanding gravitational wave progenitors and the effect of X-ray binaries in the early Universe. Studies of normal galaxies in the Chandra deep surveys shown evidence for significant cosmological evolution, which is generally attributed to the effect of metallicity. However, these studies detect only a handful of sources at the era of the peak of the star-forming activity of the universe. We discuss how different survey strategies with the ATHENA WFI will help to remedy this limitation by extending our census of normal X-ray emitting galaxies to redshifts higher than 2. A significant challenge will be the characterization of the host galaxies of the X-ray sources and the discrimination between star-forming galaxies from AGN. We discuss potential strategies for overcoming this problem by taking advantage of the available multi-wavelength data in addition to the X-ray data. Finally we discuss how the ATHENA observations fit in the current landscape of X-ray surveys (Chandra, XMM-Newton, eROSITA) and how they will help to improve our understanding of X-ray binary populations and their formation and evolution mechanisms in a cosmological context.