To investigate rates and outcomes of antibiotic de-escalation during pre-engraftment neutropenia in allogeneic hematopoietic stem cell transplantation (HSCT) recipients. 110 consecutive HSCTs performed between January 2013 and March 2014 were analyzed. De-escalation was defined as narrowing the spectrum of antibiotic treatment either within (early) or after 96 hours (late) from starting antibiotics. Discontinuation, considered a form of de-escalation, was defined as stopping antibiotics before engraftment. De-escalation failure was defined as restarting/escalating antibiotics within 96 hours after de-escalation. Predictors of de-escalation were analyzed. Among 102 patients who started antibiotics and were included, 68 (67%) received monotherapy (mainly piperacillin-tazobactam, n = 58), whereas 34 (33%) received combination therapy (mainly meropenem plus glycopeptide, n = 24). Median duration of neutropenia was 17 days. Bloodstream infections (BSIs) were diagnosed in 28 patients (20%). Early de-escalation rate was 25.5% (n = 26) and mostly consisted of reducing the spectrum of β-lactams (n = 11, 42%). In comparison with theoretical scenario of continuing therapy until engraftment, the median savings in terms of antibiotic days were 10 for meropenem, 8 for piperacillin-tazobactam, and 7 for vancomycin. Failure rate of early de-escalation was 15% (4/26). Late de-escalation rate was 30.4% (n = 31) and failure rate 19% (6/31). The rate of de-escalation any time before engraftment was 55.9% (n = 57), including discontinuation in 33 patients (32%). Death at day 60 after HSCT occurred in 3 patients who never underwent de-escalation. Acute myeloid disease and BSIs were independent predictors of early de-escalation. De-escalation, including discontinuation, is feasible and safe in pre-engraftment neutropenia after allogeneic HSCT.