Machado-Joseph disease (MJD) is a neurodegenerative disorder caused by an abnormal expansion of citosine-adenine-guanine trinucleotide repeats in the disease-causing gene. This mutation leads to an abnormal polyglutamine tract in the protein ataxin-3 (Atx3), resulting in formation of mutant Atx3 aggregates. Despite several attempts to develop a therapeutic option for MJD, currently there are no available therapies capable of delaying or stopping disease progression. Recently, our group reported that reducing the expression levels of mutant Atx3 lead to a mitigation of several MJD-related behavior and neuropathological abnormalities. Aiming a more rapid translation to the human clinics, in this study we investigate a pharmacological inhibitor of translation-cordycepin-in several preclinical models. We found that cordycepin treatment significantly reduced (i) the levels of mutant Atx3, (ii) the neuropathological abnormalities in a lentiviral mouse model, (iii) the motor and neuropathological deficits in a transgenic mouse model and (iv) the number of ubiquitin aggregates in a human neural model. We hypothesize that the effect of cordycepin is mediated by the increase of phosphorylated adenosine monophosphate-activated protein kinase (AMPK) levels, which is accompanied by a reduction in the global translation levels and by a significant activation of the autophagy pathway. Overall, this study suggests that cordycepin might constitute an effective and safe therapeutic approach for MJD, and probably for the other polyglutamine diseases.