Introduction Positive fluid balance has been associated with an increased risk for mortality in critically ill patients with acute kidney injury with or without renal replacement therapy (RRT). Data on fluid accumulation prior to RRT initiation and mortality are limited. We aimed to study the association between fluid accumulation at RRT initiation and 90-day mortality. Methods We conducted a prospective, multicenter, observational cohort study in 17 Finnish intensive care units (ICUs) during a five-month period. We collected data on patient characteristics, RRT timing, and parameters at RRT initiation. We studied the association of parameters at RRT initiation, including fluid overload (defined as cumulative fluid accumulation > 10% of baseline weight) with 90-day mortality. Results We included 296 RRT-treated critically ill patients. Of 283 patients with complete data on fluid balance, 76 (26.9%) patients had fluid overload. The median (interquartile range) time from ICU admission to RRT initiation was 14 (3.3 to 41.5) hours. The 90-day mortality rate of the whole cohort was 116 of 296 (39.2%; 95% confidence interval 38.6 to 39.8%). The crude 90-day mortality of patients with or without fluid overload was 45 of 76 (59.2%) vs. 65 of 207 (31.4%), P < 0.001. In logistic regression, fluid overload was associated with an increased risk for 90-day mortality (odds ratio 2.6) after adjusting for disease severity, time of RRT initiation, initial RRT modality, and sepsis. Of the 168 survivors with data on RRT use at 90 days, 34 (18.9%, 95% CI 13.2 to 24.6%) were still dependent on RRT. Conclusions Patients with fluid overload at RRT initiation had twice as high crude 90-day mortality compared to those without. Fluid overload was associated with increased risk for 90-day mortality even after adjustments.