Understanding the boundaries of breeding populations is of great importance for conservation efforts and estimates of extinction risk for threatened species. However, determining these boundaries can be difficult when population structure is subtle. Emperor penguins are highly reliant on sea ice, and some populations may be in jeopardy as climate change alters sea-ice extent and quality. An understanding of emperor penguin population structure is therefore urgently needed. Two previous studies have differed in their conclusions, particularly whether the Ross Sea, a major stronghold for the species, is isolated or not. We assessed emperor penguin population structure using 4,596 genome-wide single nucleotide polymorphisms (SNPs), characterized in 110 individuals (10–16 per colony) from eight colonies around Antarctica. In contrast to a previous conclusion that emperor penguins are panmictic around the entire continent, we find that emperor penguins comprise at least four metapopulations, and that the Ross Sea is clearly a distinct metapopulation. Using larger sample sizes and a thorough assessment of the limitations of different analytical methods, we have shown that population structure within emperor penguins does exist and argue that its recognition is vital for the effective conservation of the species. We discuss the many difficulties that molecular ecologists and managers face in the detection and interpretation of subtle population structure using large SNP data sets, and argue that subtle structure should be taken into account when determining management strategies for threatened species, until accurate estimates of demographic connectivity among populations can be made.