OBJECTIVE: To determine the frequency of respiratory viral infections among infants who were evaluated for late-onset sepsis in the neonatal intensive care units (NICUs) of Parkland Memorial Hospital, Dallas, Texas; and Women & Infants Hospital, Providence, Rhode Island. STUDY DESIGN: Prospective cohort study conducted from January 15, 2012 to January 31, 2013. Infants in the NICU were enrolled if they were inborn, had never been discharged home, and were evaluated for sepsis (at >72 hours of age) and antibiotic therapy was initiated. Infants had a nasopharyngeal specimen collected for detection of respiratory viruses by multiplex polymerase chain reaction within 72 hours of the initiation of antibiotic therapy. Their medical records were reviewed for demographic, clinical, radiographic, and laboratory data until NICU discharge. RESULTS: During the 13-month study, 8 of 100 infants, or 8 (6%) of the 135 sepsis evaluations, had a respiratory virus detected by polymerase chain reaction (2, enterovirus/rhinovirus; 2, rhinovirus; 2, coronaviruses; and 2, parainfluenza-3 virus). By bivariate analysis, the infants with viral detection were older (41 vs 11 days; P = .007), exposed to individuals with respiratory tract viral symptoms (37% vs 2%; P = .003), tested for respiratory viruses by provider (75% vs 11%; P < .001), and had lower total neutrophil counts (P = .02). In multivariate regression analysis, the best predictor of viral infection was the caregivers' clinical suspicion of viral infection (P = .006). CONCLUSIONS: A total of 8% of infants, or 6% of all NICU sepsis evaluations, had a respiratory virus detected when evaluated for bacterial sepsis. These findings argue for more respiratory viral testing of infants with suspected sepsis using optimal molecular assays to establish accurate diagnoses, prevent transmission, and inform antibiotic stewardship efforts.