Extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae are a common cause of bacteraemia in endemic countries and may be associated with high mortality; carbapenems are considered the drug of choice. Limited data suggest piperacillin-tazobactam could be equally effective. We aimed to compare 30-day mortality of patients treated empirically with piperacillin-tazobactam versus a carbapenem in a multi-centre retrospective cohort study in Singapore. Only patients with active empiric monotherapy with piperacillin-tazobactam or a carbapenem were included. A propensity score for empiric carbapenem therapy was derived and an adjusted multivariate analysis of mortality was conducted. A total of 394 patients had ESBL-Escherichia.coli and ESBL-Klebsiella pneumoniae bacteraemia of which 23.1% were community acquired cases. One hundred and fifty-one received initial active monotherapy comprising piperacillin-tazobactam (n = 94) or a carbapenem (n = 57). Patients who received carbapenems were less likely to have health-care associated risk factors and have an unknown source of bacteraemia, but were more likely to have a urinary source. Thirty-day mortality was comparable between those who received empiric piperacillin-tazobactam and a carbapenem (29 [30.9%] vs. 17 [29.8%]), P = 0.89). Those who received empiric piperacillin-tazobactam had a lower 30-day acquisition of multi-drug resistant and fungal infections (7 [7.4%] vs. 14 [24.6%]), P<0.01). After adjusting for confounders, use of empiric piperacillin-tazobactam was not associated with increased 30-day mortality (OR 1.00, 95% CI; 0.45-2.17). Empiric piperacillin-tazobactam was not associated with increased 30-day mortality and may result in fewer multi-drug resistant and fungal infections when compared with a carbapenem.