Abstract Tumor-associated myeloid-derived cells (MDCs) significantly impact cancer prognosis and treatment response due to their remarkable plasticity and tumorigenic behaviors. We integrated single-cell RNA-Sequencing datasets from seven different cancers, resulting in a comprehensive collection of 29 MDC subpopulations in the tumor microenvironment (TME). Distinguishing resident-tissue from monocyte-derived macrophages, we discovered a resident-tissue-like subpopulation within monocyte-derived macrophages. Additionally, hypoxia-driven macrophages emerged as a prominent TME component. Deconvolution of these profiles revealed five subpopulations as independent prognostic markers across various cancer types. Validation in large cohorts confirmed the FOLR2-expressing macrophage association with poor clinical outcomes in ovarian and triple-negative breast cancer. Moreover, the marker TREM2, commonly used to define immunosuppressive tumor-associated macrophages, cannot solely predict cancer prognosis, as different polarization states of macrophages express this marker in a context-dependent manner. This comprehensive MDC atlas offers valuable insights and a foundation for novel analyses, advancing strategies for treating solid cancers.