The current paper shows a physics informed neural network (PINN) based method to model the nonlinear magnetic force of a symmetric two pole active magnetic bearing (AMB) system. The rotor dis-placement and coil current produce the magnetic force that is a function of these quantities and the proper modeling of this is critical in high performance control. The force was analytically modeled in a way that can be described as a theoretical application of the principles of magnetic energy and finite element simulations were performed in ANSYS Maxwell to provide reference data in terms of various operating conditions. To guarantee the physical consistency, a PINN was trained on both a simulated and analytical model. The trained network was very accurate, and average variation was just 0.584 percent over FEM findings. Further, PINN performed better than a normal artificial neural network (ANN) in MATLAB that revealed dramatically greater errors in the same circumstances. The proposed model holds a great power of being incorporated into control simulations and later adaptation to experimental data that FEM cannot be applied to.