A massively parallel processor (MPP) computer has made it practical to do automatic stereo analysis of cloud-top heights from stereoscopic satellite image pairs. The automatic analysis is of equivalent quality to manual analysis while taking several orders of magnitude less time. The height of large-scale structure can be measured more accurately with the automatic analysis, but the height of small-scale towers are underestimated. Simulations using synthetic stereo data show that it is possible to automatically resolve small-scale features (e.g., 4000-m diameter clouds) to about 1500 m in the vertical. Nearly simultaneous image pairs from the GOES and NOAA satellites of hurricanes and tornadic thunderstorm clouds are used to demonstrate automatic analysis on real stereo data. Quality problems with the real stereo data require courser automatic analysis with a minimum scale size of 10 000 m. Higher quality GOES-Next stereoscopic image pairs, expected to be routinely available in the next few years should make useful operational automatic stereo cloud-height measurements practical. Some possible applications of the automatic stereo include severe storm and hurricane nowcasting, aviation forecasting and development of cloud climatologies.