Objective To identify PaCO2 trajectories and assess their associations with mortality in critically ill patients with coronavirus disease 2019 (COVID‐19) during the first and second waves of the pandemic in Denmark. Design A population‐based cohort study with retrospective data collection. Patients All COVID‐19 patients were treated in eight intensive care units (ICUs) in the Capital Region of Copenhagen, Denmark, between March 1, 2020 and March 31, 2021. Measurements Data from the electronic health records were extracted, and latent class analyses were computed based on up to the first 3 weeks of mechanical ventilation to depict trajectories of PaCO2 levels. Multivariable Cox regression analyses were used to calculate adjusted hazard ratios (aHRs) for Simplified Acute Physiology Score 3, sex and age with 95% confidence intervals (CIs) for death according to PaCO2 trajectories. Main results In latent class trajectory models, including 25,318 PaCO2 measurements from 244 patients, three PaCO2 latent class trajectories were identified: a low isocapnic (Class I; n = 130), a high isocapnic (Class II; n = 80), as well as a progressively hypercapnic (Class III; n = 34) trajectory. Mortality was higher in Class II aHR: 2.16 {1.26–3.68} and Class III aHR: 2.97 {1.63–5.40}) compared to Class I (reference). Conclusion Latent class analysis of arterial blood gases in mechanically ventilated COVID‐19 patients identified distinct PaCO2 trajectories, which were independently associated with mortality.