Properties of ground surfaces condition locomotion, and quality of track surfaces is believed to be involved in the pathogenesis of many musculoskeletal injuries in the horse. Measuring ground reaction forces (GRF) is an interesting approach to assess those interactions. Forceplates are the most commonly used but they are not well suited to compare different ground surfaces at fast gaits. Embarked equipment, fixed under the horse's hoof, would allow force measurement on any track. The objective of this work was (1) to design a device which enables the measurement of 3-D GRF on any ground, at any gait, for a given subject, (2) to determine its accuracy, and (3) to evaluate its performance and usefulness under physiological conditions. The resulting dynamometric horseshoe was composed of 4 piezoelectric sensors sandwiched between 2 aluminium plates designed at the shape of an equine shoe. The measurements, evaluated after a quasi-static calibration, revealed that the root mean square error was 1.3% in the normal direction, and 3.1% in the transversal direction. In vivo tests at the walk and trot in straight line and at the trot on circles, were conducted on 3 different ground surfaces. The results demonstrate that this dynamometric horseshoe is well suited to study the effects of different ground surfaces on GRF in the moving horse.