Persistent or recurrent unexplained fever in neutropenic patients receiving antibiotics can be caused by invasive fungal infections, which are often difficult to diagnose. Early trials of empirical antifungal therapy with amphotericin B deoxycholate (AmB) documented reductions in the frequency of and the morbidity and mortality associated with invasive fungal infections. Because of AmB's infusional and renal toxicities, subsequent trials used newer, less toxic agents, such as the lipid formulations of AmB, the extended-spectrum azoles, and, more recently, the echinocandins. To date, alternatives to AmB have shown less toxicity, but improved efficacy has been less clear. Overall, empirical antifungal therapy can help prevent the morbidity associated with many fungal infections, eliminate concerns about diagnostic pitfalls, and prevent breakthrough undetected infections. However, its potential shortcomings are overtreatment, toxicity, and increased treatment-related costs when treatment is given to persons not needing it. Newer diagnostic tools are needed to target those most in need of antifungal therapy.