The recommendation of sugarcane clones depends on several factors, as the response or performance of the clones over different cuts or harvests. The clone by harvest interaction might be difficult to identify superior clones in the final stages of the sugarcane breeding program. Thus, the objective of this study was to investigate and describe the implications of the genotype by harvest interaction in the adaptability and stability of genotypes and delineation of mega-environments from a set of multi-environment trials. Fifteen clones and four checks were evaluated in eight environments. The trait TPH (tons of pol per hectare) was evaluated in two harvests (plant cane and ratoon cane) in 2010 and 2011. The joint analysis showed significance for harvest (H), environment (E), and genotype (G) effects. The interactions GxE, ExH, GxH, and ExGxH were also significant. The last three-way interaction indicated the differential response of the genotypes over environments, and that it depends on the harvests. The overall mean of the trials was 12.77 TPH. The coefficient of variation was 8.70% and the selective accuracy was 98.63%, indicating high experimental precision. The genotypes G4, G14, and G16 were statistically superior to the check varieties used; however, these genotypes did not show high stability as described by the additive main effects and multiplicative interaction method. There was a specific adaptation between the E7 and E5 environments and the G4 and G5 genotypes, respectively. In general, the grouping of the environments was inconsistent throughout the harvests, except for the E1 and E4 environments, which exhibited similarities for the different genotypes.