AMD3100 is a CXCR4 receptor inhibitor with anti–HIV-1 activity in vitro. We tested the safety, pharmacokinetics, and antiviral effect of AMD3100 administered for 10 days by continuous intravenous infusion in an open-label dose escalation study from 2.5 to 160 μg/kg/h. Forty HIV-infected patients with an HIV RNA level >5000 copies/mL on stable antiretroviral (ARV) regimens or off therapy were enrolled. Syncytium-inducing (SI) phenotype in an MT-2 cell assay was required in higher dose cohorts. Most subjects were black (55%), male (98%), and off ARV therapy. HIV phenotype was SI (30%), non–SI (45%), or not tested (25%). One patient (5 μg/kg/h) had serious and possibly drug-related thrombocytopenia. Two patients (40 and 160 μg/kg/h) had unexpected, although not serious, premature ventricular contractions. Most patients in the 80- and 160-μg/kg/h cohorts had paresthesias. Steady-state blood concentration and area under the concentration-time curve were dose proportional across all dose levels; the median terminal elimination half-life was 8.6 hours (range: 8.1–11.1 hours). Leukocytosis was observed in all patients, with an estimated maximum effect of 3.4 times baseline (95% confidence interval: 2.9–3.9). Only 1 patient, the patient whose virus was confirmed to use purely CXCR4 and who also received the highest dose (160 μg/kg/h), had a significant 0.9-log10 copies/mL HIV RNA drop at day 11. Overall, however, the average change in viral load across all patients was +0.03 log10 HIV RNA. Given these results, AMD3100 is not being further developed for ARV therapy, but development continues for stem cell mobilization.