Background: Severe community-acquired pneumococcal pneumonia is commonly associated with bacteraemia. Although it is assumed that the bacteraemia solely derives from pneumococci entering the blood from the lungs it is unknown if other organs are important in the pathogenesis of bacteraemia. Using three models, we tested the relevance of the spleen in pneumonia-associated bacteraemia. Methods: We used human spleens perfused ex vivo to explore permissiveness to bacterial replication, a non-human primate model to check for splenic involvement during pneumonia and a mouse pneumonia-bacteraemia model to demonstrate that splenic involvement correlates with invasive disease. Findings: Here we present evidence that the spleen is the reservoir of bacteraemia during pneumonia. We found that in the human spleen infected with pneumococci, clusters with increasing number of bacteria were detectable within macrophages. These clusters also were detected in non-human primates. When intranasally infected mice were treated with a non-therapeutic dose of azithromycin, which had no effect on pneumonia but concentrated inside splenic macrophages, bacteria were absent from the spleen and blood and importantly mice had no signs of disease. Interpretation: We conclude that the bacterial load in the spleen, and not lung, correlates with the occurrence of bacteraemia. This supports the hypothesis that the spleen, and not the lungs, is the major source of bacteria during systemic infection associated with pneumococcal pneumonia; a finding that provides a mechanistic basis for using combination therapies including macrolides in the treatment of severe community-acquired pneumococcal pneumonia. Funding: Oxford University, Wolfson Foundation, MRC, NIH, NIHR, and MRC and BBSRC studentships supported the work.