This study examined the effect of submaximal and high-intensity (sprint) training on venous hemoglobin concentration ([Hb]) in Thoroughbred (Tb) racehorses before and after high-intensity exercise on the gallops. Measurements for n = 14 Tb yearlings took place during high-intensity exercise on an all-weather gallop during their breaking period before entering submaximal training (UT1) and then again as 2-year-olds before (UT2) and after sprint training (T). Weather, gallop conditions, peak heart rate (HR), peak velocity (Vpeak), and exercise distance were recorded and whole-blood jugular venous samples taken before (T0) and 5 minutes after (T5min) exercise ended for packed cell volume (PCV), [Hb], and plasma lactate concentration ([LA]) measurements. Submaximal and sprint training duration including the number of sprint training sessions were recorded for each horse and included in the analyses. Data were analyzed using repeated measures analysis of variance followed by a Bonferroni t test and Pearson correlation; significance was P ≤ .05. Environmental temperature was significantly higher with firmer footing for T versus UT1 test days, with Vpeak significantly greater for T versus UT1 horses (P = .005). T5min PCV, [Hb], and [LA] were significantly higher than T0 (P < .0001) for all training stages, with no training effects observed. Vpeak was positively correlated with duration of submaximal training (r = 0.5, P = .0009) and the number of sprint training sessions (r = 0.4, P = .005), but there were no correlations between [Hb], training, and other physiological variables. These results suggest that improvement in Vpeak is related to training duration and intensity, and is influenced by the going, but is not affected by [Hb].