Abstract Study question In couples with azoospermia undergoing intracytoplasmic sperm injection (ICSI), are neonatal outcomes different depending on sperm origin (epididymal-PESA, testicular-TESA or donor sperm)? Summary answer Newborns from PESA or TESA had similar neonatal outcomes. Comparing to donor cases, PESA/TESA cycles didn’t present increased levels of malformation or stillbirth. What is known already Since the introduction of ICSI and the demonstration that the spermatozoa derived either from the epididymis or the testis were capable of normal fertilization and pregnancy, it was possible for azoospermic men to father their own children. However, it raised concerns that the quality of spermatozoa in terms of DNA damage or maturation when collected from non-ejaculated semen could differ from that collected from ejaculated and whether sperm of different origins will affect the neonatal outcome and safety of ICSI. Study design, size, duration Retrospective cohort study performed at a reproductive medicine center, using data from 359 deliveries from 945 cycles of ICSI of couples with infertility due to azoospermia, performed between 1995 and 2021. Data were collected from electronic records. Participants/materials, setting, methods A total of 359 deliveries were divided in 3 groups, according to the sperm source: PESA (n = 138), TESA (n = 139) and DONOR (n = 82). A total of 402 newborns were evaluated: PESA (n = 162); TESA (n = 156); DONOR (n = 84). Neonatal outcomes and congenital malformations were analyzed for singletons, twins and triplets separately. Statistical analysis: Anova, chi-square and Fischer test, considering p < 0.05. Main results and the role of chance Comparing the 3 groups (PESA, TESA, and donor sperm), the following results were found: The mean birth weight (BW), length and Apgar score > 7 for singletons and twins did not differ between groups; neither did prematurity or extreme prematurity. According to birth percentile, there was a lower rate of small for gestational age (SGA) newborn in the epididymal group (PESA, 1.5% vs. TESA, 8.5% vs. sperm donor, 5.5%, p = 0.004). Also, in 158 twin births, the number of newborns with low birth weight (<2500g) was significantly smaller in the epididymal group (PESA, 32.9% vs. TESA, 62.5%; donor sperm, 73.1%, p = 0.001). There were three stillbirths in triplets (PESA group). We also found six cases of neonatal mortality (one in singleton pregnancy in TESA group, 3 in twins (two in PESA group and one donor sperm group), and two in triplets (one in TESA and one in PESA groups). The total number of congenital malformations was 8 (2.48%) (PESA n = 1, 0.62%; TESA, n = 2, 1.28%, and donor sperm, n = 5, 5.95%, p = 0.02). Limitations, reasons for caution The possibility of incomplete medical records is an important limitation to be considered in retrospective studies. Larger studies are important to better assess the risk of malformation in this population. Wider implications of the findings In this study, there were no differences on neonatal outcomes regarding the source of sperm used for ICSI, with PESA, TESA and donor groups, as well as congenital malformation. It showed equally safe options with testis, epididymal or ejaculated sperm for azoospermia treatments. Trial registration number not applicable