Genomic selection, where selection decisions are based on estimates of breeding value from genome wide-marker effects, has enormous potential to improve genetic gain in dairy and beef cattle. Although successful in dairy cattle, some major challenges remain 1) only a proportion of the genetic variance is captured, particularly for some traits 2) marker effects are rarely consistent across breeds, 3) accuracy of genomic predictions decays rapidly over time. Using full genome sequences rather than DNA markers in genomic selection could address these challenges. However, sequencing all individuals in the very large resource populations required to estimate the typically small effects of mutations on target traits would be prohibitively expensive. An alternative is to sequence key ancestors contributing most of the genetic material of the current population, and to use this reference for imputation of sequence from SNP chip data. The reference set must still be large, in order to capture for example, rare variants which are likely to explain some of the variation in our target traits. Recognising the need for a comprehensive “reference set” of key ancestors by many groups undertaking cattle research and cattle breeding programs, we have initiated the 1000 bull genomes project. The project will assemble whole genome sequences of cattle from institutions around the world, to provide an extended data base for imputation of genetic variants. This will enable the bovine genomics community to impute full genome sequence from SNP genotypes, and then use this data for genomic selection, and rapid discovery of causal mutations. Some preliminary results from the variant detection pipeline will be reported.