Breast cancer (BC) patients aged below 40 years at diagnosis experience aggressive disease and poorer survival compared to women diagnosed at 40-49 years, but the age-related biology is described to little extent. Here, we explored transcriptional alterations in BC to gain better understanding of age-related tumor biology. We studied a subset of the Bergen in-house cohort (n=127, age range 26-49) and used the NanoString Breast Cancer 360 expression panel on formalin-fixed paraffin-embedded breast cancer tissue, and publicly available global breast cancer mRNA expression data (n=204, age range 22-49), to explore differentially expressed genes between the young (age <40) and older (age 40-49) patients. Unsupervised hierarchical clustering was applied to identify gene expression-based patient clusters. We applied established computational approaches to define the PAM50 subtypes, Risk of recurrence scores (ROR), and Risk groups, and to infer the proportions of 22 immune cell types from bulk gene expression profiles of patients aged <50 at breast cancer diagnosis. Differentially expressed genes and gene sets were investigated using OncoEnrichR and g:Profiler to describe functional profiles and pathway enrichment. We identified four age-related patient clusters presenting distinct characteristics of PAM50 subtypes and ROR profiles, which demonstrated independent prognostic value when adjusted for traditional clinico-pathologic variables and the known molecular subtypes. Our findings showed better survival than expected in the basal-enriched cluster 2, and in triple negative and basal-like BC. Deconvolution analyses of immunophenotypes indicated higher levels of M0 and M1 macrophages than M2 macrophages in subsets of young BC. Our approach identifies age-based patient clusters with distinct clinico-pathologic profiles, to a large extent overlapping with the PAM50 subtypes, although with independent prognostic value in multivariate survival analyses. The patient clusters provided new insight in the immune cell distribution across tumor subtypes, potentially contributing to survival differences between the clusters and the molecular subtypes and indicating age-related mechanisms improving outcome. Our study confirms the applicability of ROR as a valid prognosticator also in a young breast cancer cohort.