To evaluate the efficiency, safety and outcome of high flow nasal cannula oxygen (HFNC) in ICU patients with acute respiratory failure. Pilot prospective monocentric study. Thirty-eight patients were included. Baseline demographic and clinical data, as well as respiratory variables at baseline and various times after HFNC initiation during 48 h, were recorded. Arterial blood gases were measured before and after the use of HFNC. Noise and discomfort were monitored along with outcome and need for invasive mechanical ventilation. HFNC significantly reduced the respiratory rate, heart rate, dyspnea score, supraclavicular retraction and thoracoabdominal asynchrony, and increased pulse oxymetry. These improvements were observed as early as 15 min after the beginning of HFNC for respiratory rate and pulse oxymetry. PaO(2) and PaO(2)/FiO(2) increased significantly after 1 h HFNC in comparison with baseline (141 ± 106 vs. 95 ± 40 mmHg, p = 0.009 and 169 ± 108 vs. 102 ± 23, p = 0.036; respectively). These improvements lasted throughout the study period. HFNC was used for a mean duration of 2.8 days and a maximum of 7 days. It was never interrupted for intolerance. No nosocomial pneumonia occurred during HFNC. Nine patients required secondary invasive mechanical ventilation. Absence of a significant decrease in the respiratory rate, lower oxygenation and persistence of thoracoabdominal asynchrony after HFNC initiation were early indicators of HFNC failure. HFNC has a beneficial effect on clinical signs and oxygenation in ICU patients with acute respiratory failure. These favorable results constitute a prerequisite to launching a randomized controlled study to investigate whether HFNC reduces intubation in these patients.