The objectives of this study were to predict most recent evaluations of young bulls entering artificial insemination (AI) sampling programs from pedigree information available at time of sampling and investigate whether prediction equations differ among AI organizations. Data were pedigree information and most recent USDA evaluations on bulls entering AI sampling programs from 1989 through 1994. Pedigree information included earliest available parent average, predicted transmitting abilities (PTA) of sire, dam, and maternal grand sire. Most recent evaluations were from May 2000 evaluations and included PTA and daughter yield deviations for milk, fat, and protein. Regression coefficients on PTA of sire and PTA of dam were less than the expected coefficient of 0.50. Accuracy of prediction as determined by R-square values was less than 12%. Inclusion of PTA of maternal grand sire after PTA of sire and dam increased the accuracy of prediction by less than 1%, but regression coefficients on PTA of maternal grand sire differed from 0. Regressions on parent average were not different among AI organizations for prediction of PTA and daughter yield deviations. Partial regression coefficients on PTA of sire differed among AI organizations for prediction of fat and protein but did not differ for milk. Coefficients on PTA of dam did not differ among organizations. These results indicate that AI organizations put different emphasis on PTA of sire in selection of sons for fat and protein.