In March 2016, Zoetis Genetics launched CLARIFIDE Plus, a product providing genomic breeding values for Holstein dairy cattle for 6 wellness traits: mastitis, metritis, retained placenta, displaced abomasum, ketosis, and lameness. To produce accurate genomic evaluations, an advanced methodology called single-step genomic best linear unbiased prediction (ssGBLUP) was implemented, which combines phenotype, pedigree, and genotype information in one single computational process. However, the current algorithm for ssGBLUP has a soft limit on the number of genotyped animals (around 150,000) due to the requirement of inverting the genomic relationship matrix. An innovative approach has been discovered and developed at University of Georgia to solve this limit issue based on the observation of a relatively small genomic dimensionality related to effective population size. The new resulting algorithm, called APY (algorithm for proven and young animals), only requires the genomic relationship matrix for a set of “proven” or “core” animals to be inverted, and memory requirement and computing cost for the majority “young” or “non-core” animals are linear. In this study, we explore the characteristics of the APY algorithm applied to about 259,000 genotyped animals currently available for evaluation of wellness traits, in terms of convergence property, computational efficiency, and consistency of genomic breeding values compared with the ones generated from a smaller data set without using APY algorithm. Results show that genomic evaluation is feasible on a very large data set with availability of the APY algorithm. Computations are efficient when implemented on a high performance computer. The required computing time ranged between about 3h to about 10h, depending on the trait, which is considered acceptable in the commercial setting. Genomic estimated breeding values (GEBV) are consistent with those obtained from a smaller data set where the regular algorithm can be implemented.