Background: Current bulk transcriptomic classification systems for bladder cancer do not consider the level of intratumor subtype heterogeneity. Objective: To investigate the extent and possible clinical impact of intratumor subtype heterogeneity across early and more advanced stages of bladder cancer. Design, setting, and participants: We performed single-nucleus RNA sequencing (RNA-seq) of 48 bladder tumors and additional spatial transcriptomics for four of these tumors. Total bulk RNA-seq and spatial proteomics data were available from the same tumors for comparison, along with detailed clinical follow-up of the patients. Outcome measurements and statistical analysis: The primary outcome was progression-free survival for non–muscle-invasive bladder cancer. Cox regression analysis, log-rank tests, Wilcoxon rank-sum tests, Spearman correlation, and Pearson correlation were used for statistical analysis. Results and limitations: We found that the tumors exhibited varying levels of intratumor subtype heterogeneity and that the level of subtype heterogeneity can be estimated from both single-nucleus and bulk RNA-seq data, with high concordance between the two. We found that a higher class 2a weight estimated from bulk RNA-seq data is associated with worse outcome for patients with molecular high-risk class 2a tumors. The sparsity of the data generated using the DroNc-seq sequencing protocol is a limitation. Conclusions: Our results indicate that discrete subtype assignments from bulk RNA-seq data may lack biological granularity and that continuous class scores may improve clinical risk stratification of patients with bladder cancer. Patient summary: We found that several molecular subtypes can exist within a single bladder tumor and that continuous subtype scores can be used to identify a subgroup of patients with poor outcomes. Use of these subtype scores may improve risk stratification for patients with bladder cancer, which can help in making decisions on treatment.