The aim of the present study was to investigate the possibility of adapting sperm cell concentration per insemination unit to the intrinsic fertility of a given bull. For this purpose, split ejaculates of seven proven Holstein sires of unknown fertility were diluted in TRIS or in Laiciphos extenders. TRIS diluted doses contained either 8.12 or 16 million total sperm cells (T8, T12, T16 respectively), while the routinely-used Laiciphos units contained 16 million sperm (L16). A total of 18.068 first artificial inseminations were performed by three different AI centers. No difference in the average fertility occurred between T16 and L16 (69.2 vs 69%), but the individual bull results varied; L16 displayed lower inter-bull variability. The nonreturn rates (60 to 90 days) obtained with T8 and T12 (65.7 and 66.8%, respectively) were significantly lower than with T16 and L16 (P < 0.02). The effect of AI center was also significant (P < 0.05). Despite a highly significant effect of the sire (P < 0.001), no interaction was found between the bull and cell concentration or the AI center. The bull effect indicates that it may be worthwile especially for the bulls of the highest genetic merit to test for the appropriate lowest number of spermatozoa per unit to allow wider commercial and genetic diffusion without compromising the average nonreturn rate of the sire.