The empirical upper luminosity boundary L_max_ of cool supergiants, often referred to as the Humphreys-Davidson limit, is thought to encode information on the general mass-loss behaviour of massive stars. Further, it delineates the boundary at which single stars will end their lives stripped of their hydrogen-rich envelope, which in turn is a key factor in the relative rates of Type-II to Type-Ibc supernovae from single star channels. In this paper we have revisited the issue of L_max_ by studying the luminosity distributions of cool supergiants (SGs) in the Large and Small Magellanic Clouds (LMC/SMC). We assemble highly-complete samples of cool SGs in each galaxy, and determine their spectral energy distributions from the optical to the mid-infrared using modern multi-wavelength survey data. We show that in both cases L_max_ appears to be lower than previously quoted, and is in the region of logL/L_{sun}_=5.5. There is no evidence for L_max_ being higher in the SMC than in the LMC, as would be expected if metallicity-dependent winds were the dominant factor in the stripping of stellar envelopes. We also show that L_max_ aligns with the lowest luminosity of single nitrogen-rich Wolf-Rayet stars, indicating of a change in evolutionary sequence for stars above a critical mass. From population synthesis analysis we show that the Geneva evolutionary models greatly over-predict the numbers of cool SGs in the SMC. We also argue that the trend of earlier average spectral types of cool SGs in lower metallicity environments represents a genuine shift to hotter temperatures. Finally, we use our new bolometric luminosity measurements to provide updated bolometric corrections for cool supergiants.