We characterise the basic far-IR properties and the gas mass fraction of massive (⟨ log(M∗/M⊙ )⟩ ≈ 11.0) quiescent galaxies (QGs) and explore how these evolve from z = 2.0 to the present day. We use robust, multi-wavelength (mid- to far-IR and sub-millimetre to radio) stacking ensembles of homogeneously selected and mass complete samples of log(M∗/M⊙ )≳ 10.8 QGs. We find that the dust to stellar mass ratio (Mdust/M∗) rises steeply as a function of redshift up to z ∼ 1.0 and then remains flat at least out to z = 2.0. Using Mdust as a proxy of gas mass (Mgas), we find a similar trend for the evolution of the gas mass fraction (fgas), with z > 1.0 QGs having fgas ≈ 7.0% (for solar metallicity). This fgas is three to ten times lower than that of normal star-forming galaxies (SFGs) at their corresponding redshift but ≳ 3 and ≳ 10 times larger compared to that of z = 0.5 and local QGs. Furthermore, the inferred gas depletion time scales are comparable to those of local SFGs and systematically longer than those of main sequence galaxies at their corresponding redshifts. Our analysis also reveals that the average dust temperature (Td) of massive QGs remains roughly constant (⟨ Td»= 21.0 ± 2.0 K) at least out to z ≈ 2.0 and is substantially colder (ΔTd ≈ 10 K) compared to that of SFGs. This motivated us to construct and release a redshift-invariant template IR SED, that we used to make predictions for ALMA observations and to explore systematic effects in the Mgas estimates of massive, high-z QGs. Finally, we discuss how a simple model that considers progenitor bias can effectively reproduce the observed evolution of Mdust/M∗ and fgas. Our results indicate universal initial interstellar medium conditions for quenched galaxies and a large degree of uniformity in their internal processes across cosmic time.