Allotetraploidization is an evolutionary innovation event for cotton genomes and is gaining increasing significance because the economical traits of polyploidy cotton have been significantly improved. However, the mechanisms responsible for sub-genome hybridization remain unclear. Here, we discuss the progression of allotetraploid cotton genome research, comparative genome studies, population genomics, the applications of genomic data, genome evolution, and the potential mechanisms underlying allotetraploidization. The first Gossypium hirsutum (Gh) cotton genome was reported in 2014; since then, updated Gh genomes and G. barbadense (Gb) genome data have been published. Thus far, all five allotetraploid cotton genomes have been sequenced and assembled. Analysis estimates that the hybridization of ancient A genomes and D genomes occurred approximately one to two million years ago (Mya). It is estimated that the formation of two economically important species (Gh and Gb) occurred approximately 8000 years ago. The fiber length and quality of Gh and Gb have been significantly improved when compared with diploid cotton G. arboreum (Ga). The newly emerged Gaussian probability density function (GPDF) algorithm revealed that two major long terminal repeat transposable element (LTR/TE) burst events occurred at 5.7 Mya and 2.0 Mya in the evolutionary history of allotetraploid cotton, and that the latter was highly correlated with the polyploidization of the A and D sub-genomes. Therefore, we suggest that large-scale TE burst events probably represent the most important evolutionary force for the formation of allotetraploid cotton genomes.