In this paper, a linear parameter-varying (LPV) control design method is evaluated experimentally on an active magnetic bearing (AMB) system. A speed-dependent LPV model of the AMB system is derived. Model uncertainties are identified using artificial neural networks, and an uncertainty weighting function is approximated for LPV control synthesis. Experiments are conducted to verify the robustness of LPV controllers for a wide range of rotational speed. This LPV control approach eliminates the need for gain-scheduling, and provides better performance than the traditional proportional-integral-derivative control for high-speed operation.