Cryptococcal meningitis is a major cause of morbidity and mortality in African AIDS patients, accounting for between 13% and 17% of deaths in Ugandan HIV-infected individuals and 44% of deaths in a cohort of HIV-seropositive South African miners. This burden of disease is a result of high incidence, especially in southern and East Africa, and high acute mortality. In much of Africa, fluconazole rather than amphotericin B was, and still is, widely used as initial therapy, for a variety of reasons. These include the availability of fluconazole through free access programmes and in generic form, and the attractiveness of an easy to use, safe oral regimen over a difficult to administer intravenous drug with significant side-effects, requiring inpatient admission and close laboratory monitoring. In addition, in the absence of antiretroviral therapy, treatment of cryptococcal meningitis has in the recent past been palliative rather than curative, reducing the rationale for more aggressive therapy, if this is associated with increased side-effects. However, what data there are suggest that outcomes with fluconazole at conventional dosage (up to 400 mg/d) as initial therapy are poor. In addition, the cost of amphotericin B, previously considerable in South Africa, has been reduced. More importantly, increasing access to antiretroviral therapy (ART) now means that the long-term prognosis of patients with cryptococcal meningitis is good, provided they survive the acute infection. We summarise the evidence that a factor contributing to high acute mortality in cryptococcal meningitis is the inadequacy of fluconazole at up to 400 mg/d as an induction regimen, and present the case for initial treatment with amphotericin B in South Africa, where feasible.