The action of zidovudine when administered to individuals with severe HIV thrombocytopenia was investigated. Four individuals with platelets < 50 x 109/l and CD4 cells > 200 x 106/l were treated with 600 mg zidovudine per day for 6 weeks, no drug for 6 weeks, 1200 mg zidovudine per day for 6 weeks, then no drug for 6 weeks. Glycocalicin, a platelet protein which correlates inversely with platelet survival, was assayed before and after treatment. Glycocalicin indices were also measured in four additional individuals with HIV thrombocytopenia. Platelet counts rose 2.5-fold [95% confidence interval (Cl), 2.0–3.0)] for four subjects who received 600 mg zidovudine per day and 4.9-fold (95% Cl, 4.0–5.8) for three subjects receiving 1200 mg zidovudine per day. Platelet counts declined during drug-free intervals. Plasma glycocalicin indices were elevated in all with untreated HIV thrombocytopenia. Indices fell after zidovudine treatment in six of seven individuals, suggesting that zidovudine prolonged platelet survival. Analysis of 170 HIV-seropositive asymptomatic individuals [mean CD4 count 474 x x 106/l, standard deviation (s.d.) 245 x 106/l] revealed that 14 (8%) had < 125 x 109/l platelets but only 2 (1%) had < 50 x 109/l platelets. Platelet counts increased spontaneously in eight individuals with mild HIV thrombocytopenia among the 10 for whom repeat counts were available.