Aims To compare steady‐state nelfinavir (NFV) pharmacokinetics in pregnant and nonpregnant HIV‐infected women. Methods Twenty‐five pregnant HIV‐infected women were selected from an ongoing observational study evaluating the pharmacokinetics of antiretroviral agents during pregnancy. Twenty of them were in the third and five in the second trimester. Data for the control group of 21 HIV‐infected nonpregnant women were taken from a previous multicentre pharmacokinetic trial. All the participating women achieved steady‐state plasma concentrations while on a highly active antiretroviral therapy (HAART) regimen including NFV (1250 mg bid) and two nucleoside reverse transcriptase inhibitors (NRTIs). Blood samples for NFV measurement were collected predose (Ctrough) and at 0.5, 1, 2, 3, 4, 5, 6, 8 and 12 h post dose. Results During the third trimester of pregnancy NFV AUC0−12 h median (range) values were 25.76 (12.61–42.74) µg h−1 ml−1, and were 32.49 (19.16–63.81) µg h−1 ml−1 in the control group mean difference − 9.30 µg h−1 ml−1; 95% confidence interval (CI) −15.76, −2.83; P < 0.05). Median oral clearance (CL/F) was significantly higher in pregnant women than in the control group (48.5 l h−1, range 29.3–99.1 l h−1 vs. 38.5 l h−1, range 19.6–65.2 l h−1; mean difference 12.6 l h−1; 95% CI 3.3, 21.9) but the difference disappeared when CL/F was adjusted for body weight. Ctrough was significantly (P < 0.01) lower in pregnant compared with nonpregnant women (median 0.8 µg ml−1, range 0–2.6 µg ml−1 vs. 1.5 µg ml−1, range 0.5–4.9 µg ml−1; mean difference −1.0 µg ml−1; 95% CI −1.7, −0.31). The median elimination half‐life of NFV observed during pregnancy was 3.7 h (range 1.4–6.6 h), compared with 5.2 (range 3.1–10.1 h) in the control group (mean difference −1.7; 95% CI −2.8, −0.51). Conclusions Our results indicate that women in the later stages of pregnancy may be exposed to subtherapeutic concentrations of NFV. Thus, adjustments in drug dosage or frequency of administration may be required.