Supermassive stars have been proposed as the progenitors of the massive ($\sim 10^{9}\,\rm{M}_{\odot}$) quasars observed at $z\sim7$. Prospects for directly detecting supermassive stars with next-generation facilities depend critically on their intrinsic lifetimes, as well as their formation rates... We use the 1D stellar evolution code Kepler to explore the theoretical limiting case of zero-metallicity, non-rotating stars, formed monolithically with initial masses between $10\,\rm{kM}_{\odot}$ and $190\,\rm{kM}_{\odot}$. We find that stars born with masses between $\sim60\,\rm{kM}_{\odot}$ and $\sim150\,\rm{kM}_{\odot}$ collapse at the end of the main sequence, burning stably for $\sim1.5\,\rm{Myr}$. More massive stars collapse directly through the general relativistic instability after only a thermal timescale of $\sim3\,\rm{kyr}$--$4\,\rm{kyr}$. The expected difficulty in producing such massive, thermally-relaxed objects, together with recent results for currently preferred rapidly-accreting formation models, suggests that such ``truly direct'' or ``dark'' collapses may not be typical for supermassive objects in the early Universe. We close by discussing the evolution of supermassive stars in the broader context of massive primordial stellar evolution and the possibility of supermassive stellar explosions. (read more)