Nosocomial coronavirus disease 2019 (COVID-19) is a major airborne health threat for inpatients. Architecture and ventilation are key elements to prevent nosocomial COVID-19 (NC), but real-life data are challenging to collect. We aimed to retrospectively assess the impact of the type of ventilation and the ratio of single/double rooms on the risk of NC (acquisition of COVID-19 at least 48 h after admission). This study was conducted in a tertiary hospital composed of two main structures (one historical and one modern), which were the sites of acquisition of NC: historical (H) (natural ventilation, 53% single rooms) or modern (M) hospital (double-flow mechanical ventilation, 91% single rooms). During the study period (1 October 2020 to 31 May 2021), 1020 patients presented with COVID-19, with 150 (14.7%) of them being NC (median delay of acquisition, 12 days). As compared with non-nosocomial cases, the patients with NC were older (79 years vs. 72 years; p < 0.001) and exhibited higher mortality risk (32.7% vs. 14.1%; p < 0.001). Among the 150 NC cases, 99.3% were diagnosed in H, mainly in four medical departments. A total of 73 cases were diagnosed in single rooms versus 77 in double rooms, including 26 secondary cases. Measured air changes per hour were lower in H than in M. We hypothesized that in H, SARS-CoV-2 transmission was favored by short-range transmission within a high ratio of double rooms, but also during clusters, via far-afield transmission through virus-laden aerosols favored by low air changes per hour. A better knowledge of the mechanism of airborne risk in healthcare establishments should lead to the implementation of corrective measures when necessary. People’s health is improved using not only personal but also collective protective equipment, i.e., ventilation and architecture, thereby reinforcing the need to change institutional and professional practices.