Background: Serum amyloid A (SAA) has been reported to hold promise as diagnostic and prognostic marker in foals. This has not been investigated thoroughly. Objectives: Evaluate admission SAA concentrations as predictor of sepsis and outcome. Animals: Five hundred and ninety hospitalized foals <14 days old. Methods: Retrospective multicenter study. Foals were scored with sepsis and survival scores, grouped according to health category (septic, sick but nonseptic, uncertain sepsis status) and outcome; septic foals were further categorized according to severity (normal sepsis, severe sepsis, and septic shock). SAA was compared between groups using Mann-Whitney test and Kruskal-Wallis test. Receiver operating characteristic curves identified optimal SAA cut off values for detecting sepsis and predicting outcome. Results: Admission SAA concentrations differed significantly between sick nonseptic foals (312.1 ± 685.4 mg/L) and septic foals (1079.7 ± 1254.5 mg/L) and increased with increasing sepsis score. SAA did not differ between sepsis severity groups. The optimal cut off for sepsis detection was 1050 mg/L (sensitivity 30.2%, specificity 90.7%). Admission SAA concentrations were lower in surviving (435.0 ± 723.6 mg/L) compared to nonsurviving foals (1062.7 ± 1440.1 mg/L) and decreased with increasing survival score. The optimal cut off for nonsurvival prediction was 1250 mg/L (sensitivity 22.1%, specificity 90.8%). Conclusions and Clinical Importance: SAA concentration was higher in septic foals and nonsurviving foals. Even though optimal cut offs for SAA to detect sepsis and predict outcome had low sensitivity, they had good specificity. SAA can therefore be used as a marker to rule out sepsis and nonsurvival.