More than 150 distinct breeds of indigenous cattle have evolved over several thousand years across the African continent. These populations present a rich phenotypic diversity encompassing Bos taurus, Bos indicus, and hybrid cattle, which reflect adaptations to a wide range of agroecological environments. Many of these breeds have evolved to persist on marginal rangelands in conditions characterized by ecophysiological and nutritional challenges, and virulent diseases and high parasite loads. The genetic basis of these adaptations to a myriad of African ecosystems, however, is poorly understood, and many cattle breeds are endangered due to uncontrolled crossbreeding and replacement by exotic breeds. Low-coverage sequencing offers the potential to improve our knowledge of the genomics diversity of African cattle, and it paves the way for cost-effective genomic selection. This technology requires high quality whole genome sequences to be merged into a panel, which is then used to impute low-coverage genomes. Accurate imputation requires a reference panel that includes genomes, which represent the genomic ancestry of the individual cattle that are to be imputed. Previous studies have shown this is difficult to achieve in African cattle due to a lack of this genomic information. To address this problem, we generated 128 high-coverage genomes from a wide range of African cattle populations of a total of 32 populations were unstudied previously. We combined these new data with over 3,300 publicly available genome datasets to develop an imputation pipeline based on this large reference panel with substantial representation of African cattle genomic diversity. Here we show that imputing downsampled high-coverage genomes recovers high genotype imputation accuracies; these are >99% for common variants (MAF >5%) and between 92% and 98% for rare variants (MAF >1%) in 0.5X coverage African and European cattle. The accuracy of our pipeline, however, was lower for Asian breeds, reflecting the poor sampling coverage in Asian countries.