Data from 3 grazing experiments in northern Australia were used in calculations of animal P balance to test the usefulness of current data in predicting P supplement requirements, and to determine what additional data need to be collected. Models were constructed of (i) gut P balance and (ii) body P balance. The gut P balance model used data on dietary and faecal P concentration, digestibility, live weight, and P supplementation rate, to estimate endogenous faecal losses of P. Endogenous loss of P, calculated assuming an absorption coefficient of 0.78, ranged from 8 to 29 mg P/kg LW/d. Endogenous losses were greater during periods of high liveweight gain and at higher rates of P intake. These trends and the range of values were similar to those from pen experiments. The body P model used the same input data to predict the changes in body P concentration (P(b)). Predicted P(b) declined during the growing season, and was either static or increased during the dry season. These seasonal trends were similar to those expected from data on the liveweight gain response to P supplement, and changes in cortical bone thickness. However, P(b) declined to levels lower than those reported in slaughter experiments. This was attributed to biases in estimating the dietary P concentration with non-resident oesophogeal-fistulated steers. P supplement requirements predicted from the relations developed in this analysis were similar to those currently recommended for growing cattle, but when extended to pregnant and lactating cows, P supplement requirements were higher than currently recommended commerically.