Knowledge of breed composition can be useful in multiple aspects of cattle production, and can be critical for analyzing the results of whole genome-wide association studies currently being conducted around the world. We examine the feasibility and accuracy of using genotype data from the most prevalent bovine genome-wide association studies platform, the Illumina BovineSNP50 array (Illumina Inc., San Diego, CA), to estimate breed composition for individual breeds of cattle. First, allele frequencies (of Illumina-defined allele B) of SNP on the array for each of 16 beef cattle breeds were defined by genotyping a large set of more than 2,000 bulls selected in cooperation with the respective breed associations to be representative of their breed. With these breed-specific allele frequencies, the breed compositions of approximately 2,000 two-, three-, and four-way cross (of 8 breeds) cattle produced at the US Meat Animal Research Center were predicted by using a simple multiple regression technique or Mendel (http://www.genetics.ucla.edu/software/mendel) and their genotypes from the Illumina BovineSNP50 array, and were then compared with pedigree-based estimates of breed composition. The accuracy of marker-based breed composition estimates was 89% when using either estimation method for all breeds except Angus and Red Angus (averaged 79%), based on comparing estimates with pedigree-based average breed composition. Accuracy increased to approximately 88% when these 2 breeds were combined into an aggregate Angus group. Additionally, we used a subset of these markers, approximately 3,000 that populate the Illumina Bovine3K (Illumina Inc.), to see whether breed composition could be estimated with similar accuracy when using this reduced panel of SNP makers. When breed composition was estimated using only SNP in common with the Bovine 3K array, accuracy was slightly reduced to 83%. These results suggest that SNP data from these arrays could be used to estimate breed composition in most US beef cattle in situations where pedigree is not known (e.g., multiple-sire natural service matings, non-source-verified animals in feedlots or at slaughter). This approach can aid analyses that depend on knowledge of breed composition, including identification and adjustment of breed-based population stratification, when performing genome-wide association studies on populations with incomplete pedigrees. In addition, SNP-based breed composition estimates may facilitate fitting cow germplasm to the environment, managing cattle in the feedlot, and tracing disease cases back to the geographic region or farm of origin.