Neurocognitive impairment (NCI) remains prevalent in HIV infection. Randomized trials have shown that physical exercise improves NCI in non-HIV-infected adults, but data on HIV-infected populations are limited. Community-dwelling HIV-infected participants (n = 335) completed a comprehensive neurocognitive battery that was utilized to define both global and domain-specific NCI. Participants were divided into "exercise" (n = 83) and "no exercise" (n = 252) groups based on whether they self-reported engaging in any activity that increased heart rate in the last 72 h or not. We also measured and evaluated a series of potential confounding factors, including demographics, HIV disease characteristics, substance use and psychiatric comorbidities, and physical functioning. Lower rates of global NCI were observed among the exercise group (15.7 %) as compared to those in the no exercise group (31.0 %; p < 0.01). A multivariable logistic regression controlling for potential confounds (i.e., education, AIDS status, current CD4+ lymphocyte count, self-reported physical function, current depression) showed that being in the exercise group remained significantly associated with lower global NCI (odds ratio = 2.63, p < 0.05). Similar models of domain-specific NCI showed that exercise was associated with reduced impairment in working memory (p < 0.05) and speed of information processing (p < 0.05). The present findings suggest that HIV-infected adults who exercise are approximately half as likely to show NCI as compared to those who do not. Future longitudinal studies might be best suited to address causality, and intervention trials in HIV-infected individuals will determine whether exercise can prevent or ameliorate NCI in this population.