Observations of high redshift quasars up to z~7 tell us that massive black holes (MBHs) were already in place, with masses well above 10^9 solar masses, when the Universe was less than 1 Gyr old. According to Soltan’s argument MBHs gain most of their mass via radiatively efficient accretion, hence we expect they formed early in the Universe as smaller seeds. To date, several formation mechanisms have been proposed, but no consensus has been reached yet. First, I will discuss the main limits of these scenarios and viable alternatives, i.e. the runaway merger of stellar mass black holes in nuclear stellar clusters and possible intermittent super-Eddington accretion phases. Secondly, I will present a study of the evolution of a quasar host candidate at z >~ 6, by means of extremely-high resolution cosmological simulations of a massive halo. Thanks to state-of-the-art sub-grid models, including a detailed treatment of the non-equilibrium chemistry of the ISM, I will compare theoretical/numerical models with current observations of z >~ 6 quasar hosts obtained by ALMA. In particular, I will show that the apparent MBH obesity observed at high redshift is due to the tracers employed for the measurements, and that this discrepancy with local observations disappears when the `true' properties of the host are used, leaving a population of `normal’ MBHs in (quickly evolving) galaxies.