The hot extrusion process of G3 nickel-base alloy was modeled and analyzed by the finite-element method (FEM) to obtain the wear depth of nodes on the die surface based on the modified Archard wear theory. BP neutral network was applied to identify the relationship between the die profile and wear depth of the extrusion die. A genetic algorithm (GA) was used to optimize the die profile which yields more uniform wear distribution on die surface. Thus, the FEM, neutral network, and genetic algorithm were combined to develop a method for the design of the optimal shape of a hot extrusion die. The result shows that the maximum wear depth of the extrusion die is decreased by about 30%, and the wear distribution along the die profile surface is more uniform, indicating that the above approach can improve the wear resistance and the life of extrusion die in hot extrusion of G3 alloy