Genomic predictions for North American Holstein bulls improve reliability and accuracy of genetic evaluations. Using genotypes from 38,416 markers and genetic evaluations from 3,576 bulls born before 1999, predictions for 1,759 bulls born between 1999 and 2002 were made. Genotypes were obtained from the Cooperative Dairy DNA Repository and the Illumina BovineSNP50 BeadChip. Genomic predictions for 5 yield traits, 5 fitness traits, 16 conformation traits, and net merit were computed using linear and nonlinear models. Nonlinear models showed higher accuracy, with R² values 0.05 to 0.38 greater than those from parent averages. Combined genomic predictions had realized reliabilities 23% greater than traditional parent averages. The greatest gains were for fat percentage due to a known gene with a large effect. Genomic predictions also increased reliability more by increasing the number of genotyped bulls than the number of markers. Genomic selection improves reliability by tracing gene inheritance even with small effects. The study found that genomic predictions outperformed traditional evaluations for all 27 traits, with gains in reliability equivalent to 11 additional daughter records. The reliability of genomic predictions increased with more genotyped bulls and more markers. The study also examined the effects of sex chromosomes and found that X chromosome effects were smaller than expected but still significant. Overall, genomic predictions significantly improved the accuracy and reliability of genetic evaluations for Holstein bulls.Genomic predictions for North American Holstein bulls improve reliability and accuracy of genetic evaluations. Using genotypes from 38,416 markers and genetic evaluations from 3,576 bulls born before 1999, predictions for 1,759 bulls born between 1999 and 2002 were made. Genotypes were obtained from the Cooperative Dairy DNA Repository and the Illumina BovineSNP50 BeadChip. Genomic predictions for 5 yield traits, 5 fitness traits, 16 conformation traits, and net merit were computed using linear and nonlinear models. Nonlinear models showed higher accuracy, with R² values 0.05 to 0.38 greater than those from parent averages. Combined genomic predictions had realized reliabilities 23% greater than traditional parent averages. The greatest gains were for fat percentage due to a known gene with a large effect. Genomic predictions also increased reliability more by increasing the number of genotyped bulls than the number of markers. Genomic selection improves reliability by tracing gene inheritance even with small effects. The study found that genomic predictions outperformed traditional evaluations for all 27 traits, with gains in reliability equivalent to 11 additional daughter records. The reliability of genomic predictions increased with more genotyped bulls and more markers. The study also examined the effects of sex chromosomes and found that X chromosome effects were smaller than expected but still significant. Overall, genomic predictions significantly improved the accuracy and reliability of genetic evaluations for Holstein bulls.