Gas foil bearings have bright application prospects in oil-free turbomachinery such as aircraft air cycle machines, compressors and gas turbines. To extend the applicability of foil bearings to high-speed and heavy-load systems, the axial force produced primarily by the pressure difference between the turbine and compressor sides must be taken into consideration. The thrust disc of the rotor is typically used to sustain the axial force and maintain the attitude of the rotor in a rotor-bearing system. In an analytical model used to predict the performance of gas foil thrust bearings (GFTBs), fluid–structure interaction must be considered because of the coupling effects of hydrodynamic lubrication and the compliance of bearing surface. In this study, a link-spring structural model was employed to calculate the equivalent vertical stiffness of the bump foil. This model exhaustively considered the effect of three factors: flexibility of the bump foil, interactions between bumps, and frictional forces at the contact surfaces. In addition, top foil deflection that directly affects film thickness is calculated using a finite-element shell model. The static and dynamic characteristics of GFTBs were predicted after solving the steady-state Reynolds equation and two linearized dynamic coefficient equations which were obtained with perturbation method by using finite-difference method. The structural deformation equation was substituted into the Reynolds equations in the calculation process. The calculated static load was compared with the published experimental data, and the results validated the theoretical analysis. Thereafter, the static and dynamic characteristics including impact of rotational speed, initial minimum film thickness and the ratio of the inlet to outlet film thicknesses were analyzed. Parametric studies were also conducted to evaluate the effect of different structural parameters of the bump foil and journal tilting on the static characteristics of GFTBs. Because the proposed analytical model is completely algebraic, it can be easily programmed for accurately and effectively investigating the performance of GFTBs.