The acute respiratory distress syndrome (ARDS) is generally a severe pulmonary disease, whose associated mortality remains high. The most severe forms of ARDS, during which the hypoxemia induced by the lung involvement is the most profound, have an even more dismal prognosis, with a mortality rate exceeding 60%, despite resorting to exceptional adjunctive therapies, like NO inhalation, prone positioning of the patients, almitrine infusion or high frequency oscillation (HFO)-type ventilation. In these situations, certain teams propose establishing an extracorporeal circuit, combining a centrifuge pump and an oxygenator membrane, to assure total pulmonary assistance (oxygenation and CO2 removal from the blood), or Extra-Corporeal Membrane Oxygenation (ECMO). The aim of ECMO is to minimize the trauma induced by mechanical ventilation and to allow the lungs to rest. Unfortunately, trials evaluating ECMO for this indication over the past few decades were failures because of the interval between the onset of the disease and the installation of assistance, the poor oxygenation and CO2-removal capacities of the devices used, and the high rate of complications linked to the apparatus (massive hemorrhages resulting from intense anticoagulation and the poor 'biocompatibility' of the circuits). However, over the past few years, decisive progress has been made in the conception and construction of ECMO circuits, rendering them more 'biocompatible', better performing and more resistant. Finally, the results of the therapeutic trial (CESAR, UK) that used the latest generation ECMO are promising. Thus, we now have strong clinical and pathophysiological rationales to evaluate, through a clinical trial with sufficient statistical power, the impact of early ECMO installation for the most severe forms of ARDS.