The giant panda was widely distributed in China and south-eastern Asia during the middle to late Pleistocene, prior to its habitat becoming rapidly reduced in the Holocene. While conservation reserves have been established and population numbers of the giant panda have recently increased, the interpretation of its genetic diversity remains controversial. Previous analyses, surprisingly, have indicated relatively high levels of genetic diversity raising issues concerning the efficiency and usefulness of reintroducing individuals from captive populations. However, due to a lack of DNA data from fossil specimens, it is unknown whether genetic diversity was even higher prior to the most recent population decline. We amplified complete cytb and 12s rRNA, partial 16s rRNA and ND1, and control region sequences from the mitochondrial genomes of two Holocene panda specimens. We estimated genetic diversity and population demography by analyzing the ancient mitochondrial DNA sequences alongside those from modern giant pandas, as well as from other members of the bear family (Ursidae). Phylogenetic analyses show that one of the ancient haplotypes is sister to all sampled modern pandas and the second ancient individual is nested among the modern haplotypes, suggesting that genetic diversity may indeed have been higher earlier during the Holocene. Bayesian skyline plot analysis supports this view and indicates a slight decline in female effective population size starting around 6000 years B.P., followed by a recovery around 2000 years ago. Therefore, while the genetic diversity of the giant panda has been affected by recent habitat contraction, it still harbors substantial genetic diversity. Moreover, while its still low population numbers require continued conservation efforts, there seem to be no immediate threats from the perspective of genetic evolutionary potential