The aim of the study was to determine whether a prior diagnostic testicle biopsy can predict success or failure of testicular sperm extraction (TESE) with intracytoplasmic sperm injection (ICSI) in patients with non-obstructive azoospermia caused by testicular failure, and what is the minimum threshold of sperm production in the testis which must be surpassed for spermatozoa to reach the ejaculate. Forty-five patients with non-obstructive azoospermia caused by testicular failure underwent diagnostic testicle biopsy prior to a planned future TESE-ICSI procedure. The diagnostic testicle biopsy was analysed quantitatively, and correlated with the quantitative findings of spermatogenesis in patients with normal spermatogenesis, as well as with the results of subsequent attempts at TESE-ICSI. Men with non-obstructive azoospermia caused by germinal failure had a mean of 0-6 mature spermatids/seminiferous tubule seen on a diagnostic testicle biopsy, compared to 17-35 mature spermatids/tubule in men with normal spermatogenesis and obstructive azoospermia. These findings were the same for all types of testicular failure whether Sertoli cell only, maturation arrest, cryptorchidism, or post-chemotherapy azoospermia. Twenty-two of 26 men with mature spermatids found in the prior testis biopsy had successful retrieval of spermatozoa for ICSI, 12 of their partners became pregnant, and are either ongoing or delivered. The study suggests that 4-6 mature spermatids/tubule must be present in the testis biopsy for any spermatozoa to reach the ejaculate. More than half of azoospermic patients with germinal failure have minute foci of spermatogenesis which are insufficient to produce spermatozoa in the ejaculate. Prior diagnostic testicle biopsy analysed quantitatively (for the presence of mature spermatids) can predict subsequent success or failure with TESE-ICSI. Incomplete testicular failure may involve a sparse multi-focal distribution of spermatogenesis throughout the entire testicle, rather than a regional distribution. Therefore, it is possible that massive testicular sampling from many different regions of the testes may not be necessary for successful TESE-ICSI.