A study was undertaken to measure the water flow (Qw) delivered by a vacuum airlift designed for recirculating aquaculture systems (RAS) in fresh (<1‰ of salinity) and sea water (35‰ of salinity). The vacuum airlift consists of two concentric tubes connected at their top to a depression chamber. The water rises in the inner tube as a result of air being injected in its lower section and flows back through the external downcomer tube. The vacuum airlift was adjusted at three different lengths: 2, 4 or 6 m and water discharge could be lifted from 0 to 30 cm. Air flow rate (Qg) varied from 0 to 80 L min−1. Different types of air injectors were tested, delivering different bubble sizes (0.1–5 mm) depending on porosity and functioning at low or high injection pressure. Results show an increase in water flow when pipe length and air flow were increased and lift height reduced. Water flow also depended on the type of water and ranged from 0 to 35 m3 h−1 (0–580 L min−1) for fresh water and only from 0 to 20 m3 h−1 (0–330 L min−1) for sea water (for a 6 m high vacuum airlift). This difference was attributed to the smaller bubble diameter and higher gas holdup (ɛg) observed in sea water (0–20%) compared to fresh water (0–10%). When bubbles were present in the downcomer tube, they created a resistance to flow (counter-current airlift) that slowed down liquid velocity and thus water flow. Increasing the vacuum made it possible to use low air injection pressures and high injection depths. Vacuum also increased bubble size and airflow (20 L min−1 at atmospheric pressure to 60 L min−1 at 0.3 barA) and thus water flow rates. With RAS, the presence of fish feed in water rapidly increased water flow delivered by the airlift because of changes of water quality and gas holdup. When working with low head RAS (under 0.3 m), vacuum airlift could save up to 50% of the energy required for centrifugal pumps. An empirical predictive model was developed and calibrated. Simulation shows a good correlation between predicted values and measurements (R2 = 0.96).