In patients infected by HIV, the efficacy of highly active antiretroviral (ARV) therapy through the blockade of different steps of the retrovirus life cycle is now well established. As HIV is a retrovirus that replicates within the cells of the immune system, intracellular drug concentrations are important to determine ARV drug efficacy and toxicity. Indeed, nucleoside reverse transcriptase inhibitors (NRTIs), non-NRTIs (NNRTIs), newly available integrase inhibitors and protease inhibitors (PIs) act on intracellular targets. NRTIs are prodrugs that require intracellular anabolic phosphorylation to be converted into their active form of triphosphorylated NRTI metabolites, most of which have longer plasma half-lives than their parent compounds. The activity of intracellular kinases and the expression of uptake transporters, which may depend on cell functionality or their activation state, may greatly influence intracellular concentrations of triphosphorylated NRTI metabolites. In contrast, NNRTIs and PIs are not prodrugs, and they exert their activity by inhibiting enzyme targets directly. All PIs are substrates of cytochrome P450 3A, which explains why most of them display poor pharmacokinetic properties with intensive presystemic first-pass metabolism and short elimination half-lives. There is evidence that intracellular concentrations of PIs depend on P-glycoprotein and/or the activity of other efflux transporters, which is modulated by genetic polymorphism and coadministration of drugs with inhibiting or inducing properties. Adequate assay of the intracellular concentrations of ARVs is still a major technical challenge, together with the isolation and counting of peripheral blood mononuclear cells (PBMCs). Furthermore, intracellular drug could be bound to cell membranes or proteins; the amount of intracellular ARV available for ARV effectiveness is never measured, which is a limitation of all published studies. In this review, we summarize the findings of 31 studies that provided results of intracellular concentrations of ARVs in HIV-infected patients. Most studies also measured plasma concentrations, but few of them studied the relationship between plasma and intracellular concentrations. For NRTIs, most studies could not establish a significant relationship between plasma and triphosphate concentrations. Only eight published studies reported an analysis of the relationships between intracellular concentrations and the virological or immunological efficacy of ARVs in HIV patients. In prospective studies that were well designed and had a reasonable number of patients, virological efficacy was found to correlate significantly with intracellular concentrations of NRTIs but not with plasma concentrations. For PIs, the only prospectively designed trial of lopinavir found that virological efficacy was influenced by both trough plasma concentrations and intracellular concentrations. ARVs are known to cause important adverse effects through interference with cellular endogenous processes. The relationship between intracellular concentrations of ARVs and their related toxicity was investigated in only four studies. For zidovudine, the relative strength of the association between a decrease in haemoglobin levels and plasma zidovudine concentrations, as compared with intracellular zidovudine triphosphate concentrations, is still unknown. Similarly, for efavirenz and neuropsychological disorders, methodological differences confound the comparison between studies. In conclusion, intracellular concentrations of ARVs play a major role in their efficacy and toxicity, and are influenced by numerous factors. However, the number of published clinical studies in this area is limited; most studies have been small and not always adequately designed. In addition, standardization of assays and PBMC counts are warranted. Larger and prospectively designed clinical studies are needed to further investigate the links between intracellular concentrations of ARVs and clinical endpoints.