The next generation of ground based telescopes will break the 20th century paradigm of the 'factor of two' diameter increase. Taking advantage of the enormous advances in technology that the present generation of 8-10m telescopes has fostered, they will be fully adaptive, fully steerable behemoths of up to 100m diameter performing at the diffraction limit in the optical and near IR. At ten times the collecting area of every telescope ever built put together, they will have limiting magnitudes of 37-38, angular resolution of 1-2 milliarcseconds, and a price tag that does not follow the historical D2.6 cost law. In this paper we discuss some of the possible science cases for a telescope of 100m. Among them the determination of H unencumbered by local effects, the study of every SN ever exploded at any z < 10, the spectroscopy of extra-solar planets, studies of ultrahigh frequency phenomena, imaging of stellar surfaces, detection of brown dwarfs in external galaxies. The advent of the next generation of Extremely Large Telescopes will probably substantially change the operation all paradigm of astronomical observations, expanding on the present trend towards large programs, much in the way particle physics has gone with the large accelerators.