The dysregulated host response to infection leading to organ dysfunction is highly heterogeneous. It is currently poorly delineated by sepsis as a clinical syndromic classification, thus confounding immunotherapy trials. Here we establish the pathophysiology and potential therapeutic targets of a specific extreme response to infection state (sepsis response signature SRS1), characterised by immune compromise and poor outcome. We first derive a whole blood single-cell multi-omic atlas of the sepsis response (2727,993 cells, n=39), finding an increase in IL1R2+ immature neutrophils in SRS1, which we confirmed by CyTOF and RNA-sequencing (n=53). We next uncovered high activity of neutrophil STAT3 gene expression programs in SRS1, which were shared across multiple infectioius disease settings (n=1044) irrespective of the clinical definition of the patient cohorts. We observed elevated plasma G-CSF and IL-6 in SRS1, suggesting heightened emergency granulopoiesis (EG). We therefore characterised patient and healthy control hematopoietic stem cells (HSCs) using single-cell RNA/chromatin accessibility multi-omics (29,366 cells, n=27), identifying SRS1-specific EG transcriptional skewing, together with STAT3 and EG master regulator CEBPB epigenetic signatures. Our findings establish a common cellular axis present across extreme responses to infection, reveal its hematopoietic origin, and nominate G-CSF and IL-6 as potential therapeutic targets for the SRS1 state. The present data deposit includes processed and quality-controlled data tables for: 1. Whole blood leukocytes profiled with the BD Rhapsody platform in a cohort of 39 sepsis patients (RNA and protein count matrices, as well as their accompanying metadata table) 2. Circulating HSCs in blood profiled with the 10X multiomics platform in a cohort of 27 sepsis patients (RNA and ATAC-seq count matrices, as well as their accompanying metadata tables)