BACKGROUND: WHO recommends referral to hospital for possible serious bacterial infection in young infants aged 0-59 days. We aimed to assess whether oral amoxicillin treatment for fast breathing, in the absence of other signs, is as efficacious as the combination of injectable procaine benzylpenicillin-gentamicin. METHODS: In a randomised, open-label, equivalence trial at five sites in DR Congo, Kenya, and Nigeria, community health workers followed up all births in the community, identified unwell young infants, and referred them to study nurses. We randomly assigned infants with fast breathing as a single sign of illness or possible serious bacterial infection, whose parents did not accept referral to hospital, to receive either injectable procaine benzylpenicillin-gentamicin once per day or oral amoxicillin treatment twice per day for 7 days. A person who was off-site generated randomisation lists using computer software. Trained health professionals gave injections, but outcome assessors were masked to group allocations. The primary outcome was treatment failure by day 8 after enrolment, defined as clinical deterioration, development of a serious adverse event including death, persistence of fast breathing on day 4, or recurrence up to day 8. The primary analysis was per protocol and we used a prespecified similarity margin of 5% to assess equivalence between regimens. This study is registered with the Australian New Zealand Clinical Trials Registry, number ACTRN12610000286044. FINDINGS: From April 4, 2011, to March 29, 2013, we enrolled 2333 infants aged 0-59 days with fast breathing as the only sign of possible serious bacterial infection at the five study sites. We assigned 1170 infants to receive injectable procaine benzylpenicillin-gentamicin and 1163 infants to receive oral amoxicillin. In the per-protocol analysis, from which 137 infants were excluded, we included 1061 (91%) infants who fulfilled predefined criteria of adherence to treatment and adequate follow-up in the injectable procaine benzylpenicillin-gentamicin group and 1145 (98%) infants in the oral amoxicillin group. In the procaine benzylpenicillin-gentamicin group, 234 infants (22%) failed treatment, compared with 221 (19%) infants in the oral amoxicillin group (risk difference -2·6%, 95% CI -6·0 to 0·8). Four infants died within 15 days of follow-up in each group. We detected no drug-related serious adverse events. INTERPRETATION: Young infants with fast breathing alone can be effectively treated with oral amoxicillin on an outpatient basis when referral to a hospital is not possible. FUNDING: Bill & Melinda Gates Foundation grant to WHO.