Purpose: To study performance, physiological and biomechanical responses during repeated upper-body sprint exercise. Methods: Twelve male elite cross-country skiers performed eight 8-s maximal poling sprints with a 22-s recovery while sitting on a modified SkiErg poling ergometer. Force, movement velocity, cycle rate, work per cycle, oxygen saturation in working muscles and pulmonary oxygen uptake were measured continuously. A 3-min all-out ergometer poling test determined VO2peak, and 1 repetition maximum (1RM) strength was determined in a movement-specific pull-down. Results: Average sprint power was 281 ± 48 W, with the highest power on the first sprint, a progressive decline in power output over the following four sprints, and a sprint decrement of 11.7 ± 4.1 %. Cycle rate remained unchanged, whereas work per cycle progressively decreased (P < 0.05). m. triceps brachii and m. latissimus dorsi were highly desaturated already after the first sprint (all P < 0.05), whereas the response was delayed for m. biceps brachii and m. vastus lateralis. Correspondingly, increases in VO2 mainly occurred over the first two sprints (P < 0.05) and plateaued at approximately 75 % of VO2peak. 1RM correlated with power during the first four sprints and with average sprint power (r = 0.71–0.80, all P < 0.05), whereas VO2peak correlated with power in the last three sprints (r = 0.60–0.71, all P < 0.05). Conclusions: The main decrement in upper-body sprint performance was evident in the first five sprints, followed by highly desaturated muscles and a plateau in pulmonary oxygen uptake already after the first 2–3 sprints. While high maximal strength seems important for producing power, aerobic capacity correlates with power in the last sprints.