Background: Upper tract urothelial carcinoma (UTUC) is an aggressive disease that is risk-stratified by clinicopathological factors due to an incomplete understanding of its molecular features. Thus, we performed transcriptomic profiling of UTUC tumors from radical nephroureterectomy specimens and compared their molecular characterization to survival outcomes. Methods: 100 UTUC tumors from 100 patients were subject to RNA sequencing and a hybridization capture-based assay for deep sequencing of cancer-associated genes, followed by unsupervised nonnegative matrix factorization clustering based on the top 10% of variant genes. Gene Set Enrichment and immune deconvolution analyses assessed for differences in the tumor microenvironments (TME) between clusters. Results: Consensus clustering analysis identified 5 biologically distinct clusters (Cluster 1 (C1) = 17, C2 = 18, C3 = 30, C4 = 11, and C5 = 24 patients), which were associated with significant differences in disease-free (DFS) (p < 0.01) and overall survival (OS) (p = 0.03). C1 and C2 were associated with pT3/4 stages and worse DFS and OS, while C5 was associated with pTa/1 stages and better DFS and OS. In terms of somatic mutation frequency differences, C3 and C4 had overall higher tumor mutation burden and mutations in epigenetic modulators, which corresponds with the transcriptomic finding of higher microsatellite instability expression signatures in these two clusters as well. Of note, all Lynch patients (N = 4) were in C3. C3 was enriched for the presence of FGFR3 driver mutations in 93% of tumors, and TP53 mutations were frequent in C2 and C4 in 47 and 55% of tumors, respectively. Differentially expressed genes and Gene Set Enrichment analyses revealed that C1 and C2 were enriched with several Hallmark inflammation signatures, such as TNF-α signaling via NF-kB, allograft rejection, inflammatory response, IL6 JAK/STAT3 signaling, and IL2 STAT5 signaling. C1 demonstrated a particularly inflammatory phenotype enriched with INF-γ and INF-α response gene sets. Lastly, in the TME deconvolution analysis, C1 and C2 had higher expression of PDL-1, immune checkpoint, immune suppression, cancer-associated fibroblasts, and myeloid inflammation surrogate signatures. C2 and C3 were enriched with CD8 T-cells, while C1 was enriched for INF-γ and hypoxia signatures. C2 had the least hypoxic TME, which may be related to stronger stromal, EMT, and angiogenesis signature signals seen. These results were then validated using an outside institution’s published cohort. Conclusions: Several differences in transcriptomic features indicate heterogeneity among UTUC tumors. Two clusters with high rates of recurrence and worse prognosis are associated with higher immune and myeloid cell infiltration. In addition to clinicopathologic factors, tumor microenvironment immune features may have potential use for disease prognostication.