The population pharmacokinetics of eltrombopag were characterized in healthy subjects (n = 111) and patients with idiopathic thrombocytopenic purpura (ITP) (n = 88) using nonlinear mixed‐effects modeling. The final model was evaluated via graphical diagnostics and through predictive check and nonparametric bootstrap procedures. A 2‐compartment model with dual sequential first‐order absorption, absorption lag time, and interoccasion variability in absorption adequately described the data. For a typical 70‐kg Caucasian male ITP patient not taking corticosteroids, estimated parameters were apparent clearance (CL/F) = 0.668 L/h, apparent volume of the central compartment (Vc/F) = 8.76 L, apparent volume of the peripheral compartment (Vp/F) = 11.3 L, and distributional clearance (Q/F) = 0.399 L/h. Eltrombopag CL/F, Vc/F, Q/F, and Vp/F increased with body weight. For the range of weights included (43–122 kg), the parameters ranged from 26% lower to 41% higher than for a 70‐kg individual. The typical eltrombopag CL/F was 33% lower in East Asians compared with other races, 26% lower in patients taking corticosteroids concomitantly, 19% lower in females compared with males, and 17% higher in healthy subjects compared with ITP patients. There was also a dose effect, with CL/F and Vc/F estimated to be respectively 68% and 55% higher for doses 20 mg or less. In conclusion, East Asian race had the largest impact on eltrombopag exposure with a lower initial dose being recommended.