The benefits of non-invasive respiratory support strategies (NIRS), such as high-flow nasal oxygen therapy, in delaying intubation remain uncertain. We used an emulated target trial approach to evaluate outcomes associated with late intubation in patients with severe acute respiratory failure due to COVID-19. We conducted a retrospective multicentre cohort study using the French prospective OUTCOMEREA database. Adult patients admitted to intensive care units (ICUs) for severe SARS-CoV-2 pneumonia with an ICU length of stay of at least seven days were included in the study. Descriptive analyses were used to compare patients who were intubated after day 6 with those who remained under NIRS. In the emulated target trial, patients who were eligible between ICU days 7 and 12 were assigned to undergo late intubation or to remain under NIRS. Weighted Cox proportional hazards models were used. The primary outcome was 60-day mortality. Of the 1206 ICU patients with SARS-CoV-2 pneumonia, 288 were still in the ICU on day 7. Of these, 65 (22.6%) subsequently underwent late intubation and had a 60-day mortality rate of 78.5%. In the emulated target trial, which included 234 patients at risk of intubation, 57 underwent late intubation. Late intubation was associated with a higher risk of death (adjusted hazard ratio: 2.89; 95% confidence interval: 1.50-3.92). The estimated absolute difference in 60-day survival was - 0.34 (95% CI - 0.62 to 0.30). The restricted mean survival time was 17.8 days shorter in the late intubation group (95% CI - 30.6 to 10.8). In this multicentre cohort, patients requiring late intubation had a very high mortality rate. In the emulated target trial analysis, late intubation was associated with poorer survival compared to continued NIRS, suggesting a subgroup of patients with progressive respiratory failure rather than a direct effect of intubation timing.