In recent years, a number of observational constraints on the AGN/galaxy co-evolution has become available, both locally (e.g. the distribution of AGN classes as a function of host stellar mass and parent dark matter halo mass) and at higher redshifts (e.g. the differential redshift evolution of the AGN luminosity function, or "AGN downsizing"): in my talk I will show that it is possible to compare those results with the predictions of theoretical models, in order to get strong insight in our understanding of these phenomena. I will first review the different theoretical approaches to the problem of co-evolution between Active galactic nuclei (AGN) and galaxies, with particular emphasis on the different paradigms describing gas accretion onto super massive black holes and their impact on the properties of their host galaxies. I will thus show that AGN feedback is nowadays a crucial ingredient for models of galaxy and AGN formation and evolution, not only to explain the redshift evolution of the AGN population but also for setting up a variety of observed host galaxy properties, such as the evolution of their stellar masses and their star formation levels. I will then extensively discuss the different "modes" of AGN feedback, their triggering mechanisms and relate them to the physical condition of the host galaxies. Finally, I will discuss the main successes and failures of the current paradigm for AGN/galaxy co-evolution and introduce possible improvements.