This study explores nighttime heat stress in two Midwestern regions in the United States, encompassing the cities of Minneapolis and Milwaukee. Daily minimum temperature data were obtained from the MACAv2-METDATA dataset at a 4-km resolution. The dataset was downloaded both for the historical (1950-2005) and Representative Concentration Pathways (RCP) 4.5 (2006-2099) simulations from 11 global climate models. MODIS land cover data at a 5´×5´ resolution were used to delineate urban and non-urban areas. Heat stress was indicated by the occurrence of hot nights in two criterions. First, the number of days with daily minimum temperatures above 300K (27°C) was counted to calculate decadal frequency. Second, the 95th percentile of daily minimum temperatures in the historical period was used as a threshold to calculate the duration of hot nights. The study finds that (1) hot nights (> 300K) are practically non-existent in the historical simulation but are likely to occur typically 2-3 times per decade with the RCP4.5 simulations; (2) the frequency of such events in the future can exceed 25 per decade in urban areas whereas it can be just about 1 per decade in non-urban areas depending on models; and (3) hot nights (> 95th percentile threshold) are likely to last longer in the future simulations. Overall, heat stress is projected to increase both in frequency and duration, and the urban heat island effect in terms of heat stress is projected to intensify in the future.