The objectives of this study were to compare the effects of post-ruminal and intravenous infusions of wheat starch or glucose (CHO) or a mixture of amino acids (AA) on milk protein yield, nitrogen utilisation, plasma metabolites and mammary extraction rate of dairy cows in early lactation. Eight cow, ruminally fistulated, was assigned to two 4 £ 4 Latin squares during 14-day periods, where the last 7 days were for infusions. Infusions were: (1) starch in the abomasum (SP), (2) glucose in the blood (GB), (3) AA in the abomasum (AP), and (4) AA in the blood (AB). The experiment started 54^4 days (mean^S.E.) post partum (milk yield 33.4^1.7 kg). Daily amounts of nutrients infused were 378, 365, 341, and 333 g for SP, GB, AP and AB, respectively. The cows were fed a basal diet consisting of a concentrate mixture and grass silage (55:45 on dry-matter (DM) basis), and DM intake was 17.2 kg/day. Milk production was affected by site of infusion within substrate, whereas infusion substrates within infusion site (CHO or AA) were of minor importance. Compared with SP infusion, GB infusion increased ( P , 0.05) milk protein yield and concentration by 55 g and 1 g/kg. The AB infusion tended to ( P , 0.10) increase milk yield and ECM and increased ( P , 0.05) protein yield and concentration by 1.8 and 2.2 kg, 83 g and 1.1 g/kg compared with AP infusion, respectively. Nitrogen balance data indicated higher losses of metabolic faecal nitrogen (MFN) by abomasal than by intravenous infusions, and an increased ( P , 0.05) catabolism for AP and AB infusions compared with SP and GB infusions. GB infusion did not increase ( P . 0.10) plasma glucose or insulin concentrations above that of SP infusion. Compared with the SP infusion, the GB infusion had minor effect on plasma AA. AP infusion increased ( P , 0.05) plasma non-essential AA (NEAA) concentration compared with AB infusion, whereas infusion site of AA had no effect ( P . 0.05) on essential AA (EAA) or branched-chain AA (BCAA). Although a higher milk protein synthesis was observed for AB infusion, the mammary extraction rate was not higher ( P . 0.05) than for AP infusion. Across infusion site, AP and AB infusions increased plasma concentration of EAA and BCAA, but compared with GB infusion, the mammary extraction rates tended ( P , 0.10) to be lower. It is concluded that abomasal nutrient infusion increases loss of MFN and that the gastrointestinal metabolism influences the nutrients available for milk synthesis. Our conclusion is that when glucose was infused, AA limited a further milk protein synthesis, but when AA was infused, glucose or energy substrate might have been the limiting factor. Our results verify that glucogenic substrates are limiting when cows are in negative energy balance.