Purpose Intubation in patients with respiratory failure can be avoided by high-flow nasal cannula (HFNC) use. However, it is unclear whether waiting until HFNC fails, which would delay intubation, has adverse effects. The present retrospective observational study assessed overall ICU mortality and other hospital outcomes of patients who received HFNC therapy that failed. Methods All consecutive patients in one tertiary hospital who received HFNC therapy that failed and who then required intubation between January 2013 and March 2014 were enrolled and classified according to whether intubation started early (within 48 h) or late (at least 48 h) after commencing HFNC. Results Of the 175 enrolled patients, 130 (74.3 %) and 45 (25.7 %) were intubated before and after 48 h of HFNC, respectively. The groups were similar in terms of most baseline characteristics. The early intubated patients had better overall ICU mortality (39.2 vs. 66.7 %; P = 0.001) than late intubated patients. A similar pattern was seen with extubation success (37.7 vs. 15.6 %; P = 0.006), ventilator weaning (55.4 vs. 28.9 %; P = 0.002), and ventilator-free days (8.6 ± 10.1 vs. 3.6 ± 7.5; P = 0.011). In propensity-adjusted and -matched analysis, early intubation was also associated with better overall ICU mortality [adjusted odds ratio (OR) = 0.317, P = 0.005; matched OR = 0.369, P = 0.046]. Conclusions Failure of HFNC might cause delayed intubation and worse clinical outcomes in patients with respiratory failure. Large prospective and randomized controlled studies on HFNC failure are needed to draw a definitive conclusion.