The conventional wisdom is that long-wavelength (>100 micron) emission from cold (T ~ 20-40 K) dust is heated exclusively by emission from young, massive stars; consequently, the far-IR emission at long wavelengths can be used as a "pure" tracer of star formation even when the bolometric luminosity of the AGN is much greater than that of the host galaxy's stellar population. Recently, a number of groups have revisited this topic by performing radiative transfer on hydrodynamical simulations of AGN-hosting galaxies. I will present simulation results that demonstrate that contrary to the conventional wisdom, obscured AGN can drive cold dust emission on kpc scales, sometimes dominating the dust heating. Consequently, the FIR luminosity may overestimate the star formation rate of the most dust-obscured AGN-hosting galaxies, and high-redshift IR-luminous galaxies may harbor more obscured AGN than we think.