Earlier studies with temporarily isolated rumen of heifers show saturation kinetics of Mg efflux across the rumen wall. Therefore, we hypothesized that high Mg intakes would not further increase the rate of Mg absorption in cows. To test our hypothesis, six ruminally fistulated non-pregnant dry cows were given diets with different Mg concentrations in a 6×6 Latin square design. Desired concentrations of Mg were attained by adding MgO to the basal diet and the Mg concentrations in the total rations were 3·8, 6·4, 9·1, 11·8, 14·1 and 17·3 g Mg/kg dry matter, which provided Mg intakes of 27·1, 44·6, 64·6, 83·5, 100·4 and 124·3 g/d, respectively. Increasing Mg intakes were associated with increased (P<0·001) faecal Mg excretion. However, apparent Mg absorption expressed as g/d was not significantly different for Mg intakes from 100·4 to 124·3 g/d while Mg absorption expressed as a proportion of intake was not significantly different for Mg intakes ranging from 64·6 to 124·3 g/d. Mg concentrations in rumen fluid after feeding increased (P<0·001) with increasing Mg intakes. Apparent absorption of Mg appeared to become saturated at a ruminal Mg concentration of 17·5 mM (Mg intake of 83·5 g/d). Group-mean post-feeding concentrations of Mg and Na in rumen fluid were significantly correlated (Pearson's r=−0·96; P=0·003, n=6). This study showed that under conditions of practical dairy cow feeding, Mg absorption was maximal at Mg intakes [ges ]84 g/d.