There is now incontrovertible evidence that a combination of two nucleoside analogues is superior to AZT monotherapy for the treatment of HIV infection. There is also evidence that a range of antiretroviral combinations can produce profound drops in plasma HIV viral load, often below the detectable limits of the assay. A lower rate of viral replication is associated with delays in the selection of viral mutants resistant to drugs; the principal reason for therapeutic failure. These highly effective combinations are likely to delay clinical events and considerably improve the prognosis in HIV-infected individuals. Using mathematical modelling techniques to assess the half-life of plasma virus infectious cells and proviral DNA, one group has gone so far as to suggest it may be possible to stop therapy after several years, as the virus will have been eliminated. More sanguine investigators, however, believe that this is unlikely, as some infected cellular compartments may have long half-lives and the virus may persist, or even continue to replicate slowly, in sanctuary sites, such as the CNS and genital tract, poorly penetrated by drugs. Thus, some virologists and clinicians believe that it is crucially important to give initial therapy with the strongest possible combination of drugs to achieve an 'undetectable' plasma viral load and that, providing compliance is good, this will lead to a prolonged therapeutic effect. In these circumstances no sequencing issues arise. Others, however, continue to believe that, in due course, the virus is likely to evade drug pressure. In these circumstances, consideration of a sequence of drugs, to avoid initial and subsequent therapy from squandering future therapy options through, principally, cross-resistance, becomes an important issue. None of the issues discussed in this review has been tested by controlled clinical trials but they generate hypotheses requiring urgent evaluation.