This study was conducted to evaluate the effect of phosphorus (P) supplementation, with or without addition of other minerals, ionophore and antibiotic, on performance, P utilisation and profitability in finishing Nellore bulls in a feedlot. Forty-two animals, with an initial bodyweight of 296 ± 25 kg, were distributed in a completely randomised design consisting of three treatments and 14 replicates. The treatments were without P supplementation (T1), commercial mineral supplement specific for feedlot cattle (T2) and supplementation with dicalcium phosphate (T3). The T3 was formulated to achieve a P concentration similar to that of T2, but without adding other additives (e.g. vitamins, ionophores, yeast and chelates) in the mineral mix. The P concentrations were 2.4, 4.2 or 5.0 g P/kg of dry matter (DM), respectively, for T1, T2 and T3. The diet, on a DM basis, comprised sugarcane bagasse (200 g/kg) plus concentrate (800 g/kg). The DM, crude protein and neutral detergent fibre intakes and apparent DM digestibility were similar among the three treatments (P > 0.05). Despite the numerical difference on P concentration in T2 and T3 treatments, P intakes were similar in these treatments. The addition of commercial mineral supplement specific for feedlot cattle or dicalcium phosphate in diets increased P excretion in the faeces and urine, increased P retention and reduced the profitability of the feedlot. Performance and carcass traits were not affected by diet P concentrations (P > 0.05). The profit per head in T1 treatment was US$13.7 and the addition of the dicalcium phosphate and mineral mix in T2 and T3 did not result in profit, these treatments gave economic losses of US$6.80 and US$6.20 per head, respectively. Under Brazilian conditions, feedlot Nellore cattle fed high-grain diets do not require any additional mineral supplements. P concentration of 2.4 g/kg DM, as used in the control diet, was adequate to ensure animal performance and reduce faecal P excretion, which is in agreement with NRC and CSIRO recommendations. Lower values may also be appropriate, but were not tested here.