BACKGROUND: Parenteral antibiotic therapy for young infants (aged 0-59 days) with suspected sepsis is sometimes not available or feasible in countries with high neonatal mortality. Outpatient treatment could save lives in such settings. We aimed to assess the equivalence of two simplified antibiotic regimens, comprising fewer injections and oral rather than parenteral administration, compared with a reference treatment for young infants with clinical severe infection. METHODS: We undertook the Simplified Antibiotic Therapy Trial (SATT), a three-arm, randomised, open-label, equivalence trial in five communities in Karachi, Pakistan. We enrolled young infants (aged 0-59 days) who either presented at a primary health-care clinic or were identified by a community health worker with signs of clinical severe infection. We included infants who were not critically ill and whose family refused admission. We randomly assigned infants to either intramuscular procaine benzylpenicillin and gentamicin once a day for 7 days (reference); oral amoxicillin twice daily and intramuscular gentamicin once a day for 7 days; or intramuscular procaine benzylpenicillin and gentamicin once a day for 2 days followed by oral amoxicillin twice daily for 5 days. The primary outcome was treatment failure within 7 days of enrolment and the primary analysis was per protocol. We judged experimental treatments as efficacious as the reference if the upper bound of the 95% CI for the difference in treatment failure was less than 5·0. This trial is registered at ClinicalTrials.gov, number NCT01027429. FINDINGS: Between Jan 1, 2010, and Dec 26, 2013, 2780 infants were deemed eligible for the trial, of whom 2453 (88%) were enrolled. Because of inadequate clinical follow-up or treatment adherence, 2251 infants were included in the per-protocol analysis. 820 infants (747 per protocol) were assigned the reference treatment of procaine benzylpenicillin and gentamicin, 816 (751 per protocol) were allocated amoxicillin and gentamicin, and 817 (753 per protocol) were assigned procaine benzylpenicillin, gentamicin, and amoxicillin. Treatment failure within 7 days of enrolment was reported in 90 (12%) infants who received procaine benzylpenicillin and gentamicin (reference), 76 (10%) of those given amoxicillin and gentamicin (risk difference with reference -1·9, 95% CI -5·1 to 1·3), and 99 (13%) of those treated with procaine benzylpenicillin, gentamicin, and amoxicillin (risk difference with reference 1·1, -2·3 to 4·5). INTERPRETATION: Two simplified antibiotic regimens requiring fewer injections are equivalent to a reference treatment for young infants with signs of clinical severe infection but without signs of critical illness. The use of these simplified regimens has the potential to increase access to treatment for sick young infants who cannot be referred to hospital. FUNDING: The Saving Newborn Lives initiative of Save the Children, through support from the Bill & Melinda Gates, and by WHO and USAID.