Fatigue is among the most common symptoms reported by HIV-infected individuals. Previous reports suggest that the prevalence of fatigue varies by disease status with rates close to 80% in patients with AIDS. However, most studies have not been conducted in the setting of a controlled trial and have not assessed the association of fatigue with cellular markers of brain activity. Data for this study were derived from baseline and longitudinal evaluations in ACTG A5090, a randomized, double-blind, placebo-controlled trial of the Selegiline Transdermal System for the treatment of HIV-associated cognitive impairment. Fatigue was assessed using the Fatigue Severity Scale with scores of >4 considered "fatigued". Participants in a substudy underwent brain magnetic resonance spectroscopy (MRS) imaging, an in vivo method for assessing brain metabolites associated with neuronal and glia activity. Differences between fatigued and non-fatigued participants were evaluated with respect to demographics and clinical characteristics, plasma and CSF HIV-1 RNA concentration, CD4 counts, and brain metabolites. One hundred and twenty-eight participants were enrolled (88% male, median age = 45 years) and 82 participants (64%, 95% confidence interval 55%, 72%) were fatigued at baseline. MRS was conducted in 62 of the 128 participants. Fatigued participants were significantly younger (p = 0.011), had lower Karnofsky scores (p = 0.032), and had higher levels of depressive symptoms on the Center for Epidemiologic Studies Depression (CES-D) scale (p < 0.001) than non-fatigued participants. Statistically significant differences between fatigued and non-fatigued groups were not detected for plasma and CSF HIV-1RNA concentration, CD4 counts, or on neuropsychological tests. MRS revealed significantly lower levels of the cellular energy marker total creatine (p = 0.002) in the basal ganglia of fatigued participants. Statistically significant differences in other brain metabolites were not detected. Longitudinal data showed that fatigue persisted and worse fatigue at baseline was predictor of future fatigue. Among the cognitive tests, baseline Stroop score was associated with future fatigue. Fatigue was present in 64% of A5090 study participants and persisted during the 24 weeks of follow-up. Fatigue was associated with worse functional performance and depressive mood. Lower cellular energy levels in the basal ganglia, as measured by MRS total creatine concentration, suggest energy dysmetabolism in this brain region. This observation, taken together with the association between fatigue and neuropsychological tests of frontal lobe performance is consistent with the hypothesis of a striatal-cortical circuitry involvement in the symptoms of fatigue.