Background: Hemorrhagic stroke (HS) accounts for approximately 30% of all stroke cases and has high mortality in the intensive care unit (ICU). The blood urea nitrogen-to-creatinine ratio (BUNCR) is a potential biomarker of catabolic stress in critically ill patients. Meanwhile, its dynamic prognostic value in ICU-admitted HS patients remains unclear. This study utilized Group-based Trajectory Modeling (GBTM) to investigate associations between early BUNCR patterns and mortality. Methods: This study was conducted using data from the MIMIC-IV (v2.2) database. HS cases were identified via ICD-9/10. BUNCR trajectories were assessed by applying GBTM during the first 7 days. Outcomes were all-cause mortality (ACM) on day 28, on day 90, and at 1 year, with ICU and in-hospital mortality also evaluated. Kaplan–Meier survival curves and log-rank test compared survival across groups. Multivariable Cox proportional models adjusted for confounders and subgroup analysis assessed robustness. Results: Among 2559 patients (52.48% male), mortality was 10.16% (ICU), 14.07% (in-hospital), 17.00% (28-day), 22.43% (90-day), and 36.97% (1-year). Three BUNCR trajectories were identified: Group 1 (upward–downward, n = 655), Group 2 (stable upward, n = 1270), and Group 3 (downward–upward, n = 634). Group 2 had the highest ACM risk at 28-day, 90-day, and 1-year (p < 0.01), and was identified as a significant risk factor in multivariate Cox regression. Subgroup revealed significant interactions of BUNCR trajectories with age and sepsis. Conclusions: Distinct BUNCR trajectories were significantly associated with ACM in critically ill HS patients. Persistently increasing BUNCR predicted the poorest outcomes, underscoring its potential as a dynamic biomarker for timely risk stratification and informed ICU decisions.