Abstract Background Tuberculosis/HIV co-infection is a bidirectional and synergistic combination of two very important pathogens in public health. To date, there have been limited clinical data regarding mortality rates among tuberculosis/HIV co-infected patients and the impact of antiretroviral therapy on clinical outcomes in Ethiopia. This study assessed the incidence and predictors of tuberculosis/HIV co-infection mortality in Southwest Ethiopia. Methods A retrospective cohort study collated tuberculosis/HIV data from Jimma University Teaching Hospital for the period of September 2010 and August 2012. The data analysis used proportional hazards cox regression model at P value of ≤ 0.05 in the final model. Results Fifty-five (20.2 %) patients died during the study period and 272 study participants contributed 3 082.7 person month observations. Factors including: being aged between 35–44 years (AHR = 2.9; 95 % CI : 1.08–7.6), being a female sex worker (AHR = 9.1; 95 % CI : 2.7–30.7), being bed ridden as functional status (AHR = 3.2; 95 % CI : 1.2–8.7), and being at World Health Organization HIV disease stages 2 (AHR = 0.2; 95 % CI : 0.06–0.5), 3(AHR = 0.3; 95 % CI : 0.1–0.8) and 4(AHR = 0.2; 95 % CI : 0.04–0.55) were significant predictors of mortality for tuberculosis/HIV co-infected patients. Conclusions Contrary to our expectations, the World Health Organization (WHO) HIV disease stage 1 was found to be a significant predictor of mortality. Higher mortality rates were observed in WHO disease stage 1 patients compared to patients in stages 2, 3 and 4. The current study also confirmed and reaffirmed known significant predictors of the mortality for tuberculosis/HIV co-infected patients including being 35–44 years, being a female sex worker and being bed ridden functional status. The occurrence of high death rate among tuberculosis/HIV co-infected cases needs actions to reduce this poor outcome.