We analysed a representative collection of New World sweet potato landraces (329 accessions from Mexico to Peru) with both chloroplast and nuclear microsatellite markers. Both kinds of markers supported the existence of two geographically restricted genepools, corresponding to accessions from the north-western part of South America and accessions from the Caribbean and Central America region. Our conservative cpSSRs markers revealed that the divergence between the two haplotype groups is associated with numerous mutation events concerning various markers, supporting the idea that this divergence may be ancient, predating domestication. For both kinds of markers, we found no significant difference in diversity between the two genepools and detected region-specific alleles in both groups. Previous studies have favoured the hypothesis of a single domestication of this crop. Our analysis suggests at least two independent domestications, in Central/Caribbean America and in the north-western part of South America. Sweet potato was then dispersed from these centres throughout tropical America. Comparison of nuclear and chloroplast data suggests that exchanges of clones and sexual reproduction were both important processes in landrace diversification in this clonally propagated crop. Our analysis provides useful tools for rationalizing the conservation and use of sweet potato germplasm collections.