ABSTRACT One of the most air-reliant obligate air-breathing fish is the South American Arapaima gigas, with substantially reduced gills impeding gas diffusion, thought to be a result of recurring aquatic hypoxia in its habitat. In normoxic water, A. gigas is reported to satisfy 70–80% of its O2 requirement from the air while excreting 60–90% of its CO2 to the water. If this pattern of gas exchange were to continue in severely hypoxic water, O2 loss at the gills would be expected. We hypothesized therefore that partitioning of CO2 would shift to the air phase in severe aquatic hypoxia, eliminating the risk of branchial O2 loss. By adapting a respirometer designed to measure aquatic ṀO2/ṀCO2, we were able to run intermittent closed respirometry on both water and air phase for both of these gasses as well as sample water for N-waste measurements (ammonia-N, urea-N) so as to calculate metabolic fuel utilization. In contrast to our prediction, we found that partitioning of CO2 excretion changed little between normoxia and severe hypoxia (83% versus 77% aquatic excretion, respectively) and at the same time there was no evidence of branchial O2 loss in hypoxia. This indicates that A. gigas can utilize distinct transfer pathways for O2 and CO2. Routine and standard ṀO2, N-waste excretion and metabolic fuel utilization did not change with water oxygenation. Metabolism was fuelled mostly by protein oxidation (53%), while carbohydrates and lipids accounted for 27% and 20%, respectively.