Analysis of genotype (G)-by-environment (E) interactions and their influence on performance of cotton (Gossypium hirsutum L.) cultivars can help cotton breeders improve performance stability of cultivars across environments. Data from multi-location trials of the Regional High Quality Tests conducted as part of the USDA-ARS National Cotton Variety Tests during 2003 and 2009 were used to analyze G × E and relationships among test locations for megaenvironments. The trials were located in the Western, Plains, Central, Delta, and Eastern regions of the U.S. Cotton Belt. Effects of G × location for lint yield were either larger or comparable to the effects of G × year. The relationships among test locations were analyzed in GGE biplot and no clear megaenvironments were identified among test locations across years. Nevertheless, the locations of Las Cruces, NM in the Western and Lubbock, TX in the Plains test regions were identified as distinct from the test locations in the other areas. It was hypothesized that the environments in the U.S. Cotton Belt belonged to one megaenvironment with the areas in the Western and Plains as a subregion. The daily minimum temperature was significantly correlated to environment scores of the first principal component axis with r values -0.41 and -0.30 for the early and late growing seasons, respectively. This result suggests that genetic improvement of cotton cultivars for tolerance to low temperature during the early and late growing season could increase yield stability.