Heat stress (HS) decreases milk protein synthesis beyond what would be expected based on the concomitant reduction in feed intake. The aim of the present study was to evaluate the direct effects of HS on milk protein production. Four multiparous, lactating Holstein cows (101 ± 10 d in milk, 574 ± 36 kg of body weight, 38 ± 2 kg of milk/d) were individually housed in environmental chambers and randomly allocated to 1 of 2 groups in a crossover design. The study was divided into 2 periods with 2 identical experimental phases (control phase and trial phase) within each period. During phase 1 or control phase (9 d), all cows were housed in thermal neutral conditions (TN; 20°C, 55% humidity) and fed ad libitum. During phase 2 or treatment phase (9 d), group 1 was exposed to cyclical HS conditions (32 to 36°C, 40% humidity) and fed ad libitum, whereas group 2 remained in TN conditions but was pair-fed (PFTN) to their HS counterparts to eliminate the confounding effects of dissimilar feed intake. After a 30-d washout period in TN conditions, the study was repeated (period 2), inverting the environmental treatments of the groups relative to period 1: group 2 was exposed to HS and group 1 to PFTN conditions. Compared with PFTN conditions, HS decreased milk yield (17.0%), milk protein (4.1%), milk protein yield (19%), 4% fat-corrected milk (23%), and fat yield (19%). Apparent digestibility of dry matter, organic matter, neutral detergent fiber, acid detergent fiber, crude protein, and ether extract was increased (11.1-42.9%) in HS cows, as well as rumen liquor ammonia (before feeding 33.2%; after feeding 29.5%) and volatile fatty acid concentration (45.3%) before feeding. In addition, ruminal pH was reduced (9.5 and 6% before and after feeding, respectively) during HS. Heat stress decreased plasma free amino acids (AA; 17.1%) and tended to increase and increased blood, urine, and milk urea nitrogen (17.2, 243, and 24.5%, respectively). Further, HS cows had reduced plasma glucose (8%) and nonesterified fatty acid (39.8%) concentrations compared with PFTN controls. These data suggest that HS increases systemic AA utilization (e.g., decreased plasma AA and increased nitrogen excretion), a scenario that limits the AA supply to the mammary gland for milk protein synthesis. Furthermore, the increase in AA requirements during HS might represent the increased need for gluconeogenic precursors, as HS is thought to prioritize glucose utilization as a fuel at the expense of nonesterified fatty acids.