Abstract Background Mother-to-child transmission of HIV (MTCT) accounts for more than 90% of pediatric Acquired Immunodeficiency Syndrome (AIDS) cases. Prevention of mother to child transmission (PMTCT) programs are provided for dual benefits i.e. prevention of HIV transmission from mother to child and enrolment of infected pregnant women and their families into antiretroviral treatment (ART). This study assessed risk and predictors of HIV transmission among HIV-exposed infants on follow up at a PMTCT clinic in a referral hospital. Methods Institution based retrospective follow up study was carried at Gondar University referral hospital PMTCT clinic. All eligible records of HIV-exposed infants enrolled between September 2005 and July 2011 were included. A midwife nurse collected data using a structured data extraction format. Data were then entered in to EPI INFO Version 3.5.1 statistical software and analyzed by SPSS version 20.0. Both bivariate and multivariate analyses were carried out to identify associations. Results A total of 509 infant records were included in the analysis. The median age of infants at enrolment to follow up was 6 weeks (inter quartile range [IQR] = 2 weeks). A total of 51(10%, 95% CI: 7.8% - 13%) infants were infected with HIV. Late enrolment to the exposed infant follow up clinic (Adjusted Odds Ratio [AOR] = 2.89, 95% CI: 1.35, 6.21), rural residence (AOR = 5.05, 95% CI: 2.34, 10.9), home delivery (AOR = 2.82, 95% CI: 1.2, 6.64), absence of maternal PMTCT interventions (AOR = 5.02, 95% CI: 2.43, 10.4) and mixed infant feeding practices (AOR = 4.18, 95% CI: 1.59, 10.99) were significantly and independently associated with maternal to child transmission of HIV in this study. Conclusions There is a high risk of MTCT of HIV among exposed infants on follow up at the PMTCT clinic of the University of Gondar referral hospital. The findings of this study will provide valuable information for policy makers to enhance commitment and support for rural settings in the PMTCT scaling-up program.