Résumé : The identification of red, apparently massive galaxies at $z>7$ in early JWST photometry suggests a strongly accelerated timeline compared to standard models of galaxy growth. A major uncertainty in the interpretation is whether the red colors are caused by evolved stellar populations, dust, or other effects such as emission lines or AGN. Here we show that three of the massive galaxy candidates at $z=6.7-8.4$ have prominent Balmer breaks in JWST/NIRSpec spectroscopy from the RUBIES program. The Balmer breaks demonstrate unambiguously that stellar emission dominates at $\lambda_{\rm rest} = 0.4\,\mu$m, and require formation histories extending hundreds of Myr into the past in galaxies only 600--800 Myr after the Big Bang. Two of the three galaxies also show broad Balmer lines, with H$\beta$ FWHM $>2500~{\rm km\,s^{-1}}$, suggesting that dust-reddened AGN contribute to, or even dominate, the SEDs of these galaxies at $\lambda_{\rm rest}\gtrsim 0.6\,\mu$m. All three galaxies have relatively narrow [O III] lines, seemingly ruling out a high-mass interpretation if the lines arise in dynamically-relaxed, inclined disks. Yet, the inferred masses also remain highly uncertain. We model the high-quality spectra using Prospector to decompose the continuum into stellar and AGN components, and explore limiting cases in stellar/AGN contribution. This produces a wide range of possible stellar masses, spanning $M_\star \sim 10^9 - 10^{11}\,{\rm M_{\odot}}$. Nevertheless, all fits suggest a very early and rapid formation, most of which follow with a truncation in star formation. Potential origins and evolutionary tracks for these objects are discussed, from the cores of massive galaxies to low-mass galaxies with over-massive black holes. Intriguingly, we find all of these explanations to be incomplete; deeper and redder data are needed to understand the physics of these systems.