Introduction: The type of admission to the Intensive Care Unit (ICU) influences the prognosis of patients with severe pneumonia and, in the case of patients with COVID-19 pneumonia, this is still unexplored. The objective of this study was to determine the differences between early and late ICU admission. Methodology: A retrospective cohort study of patients with COVID-19 pneumonia at two high-complexity hospitals in Colombia. Early ICU admission (EICUA) was defined as direct admission from the emergency department or within the first 24 hours of admission. Late ICU admission (LICUA) was defined as admission from the hospitalization service after 24 hours of arrival. A robust Cox regression was performed for the variable recovery time, to determine the impact of the ICU admission type in the hazard rate. Results: 68.2% were EICUA patients and 31.8% were LICUA patients. Recovery and duration of hospital stay were significantly lower in EICUA than in LICUA (9 vs 15 days, p = 0.0001, and 10 vs 15.5 days, p < 0.0001, respectively). However, the duration of ICU stay (7 vs 9 days, p = 0.131) and the invasive mechanical ventilation requirement (48.9% vs 54.9%, p = 0.374) were not statistically significant. The 30-day follow-up showed no difference between the EICUA and LICUA (alive 97% vs 94.6%, p = 0.705). Conclusions: Mortality between EICUA and LICUA patients with COVID-19 pneumonia showed no statistically significant differences. However, the recovery time, the probability intensity of instant recovery, and the duration of hospital stay were better in EICUA than in LICUA. Neither EICUA nor LICUA affects the final status (death) of patients.