Abstract: This paper investigates the axial compressive performance of GFRP Confined Hollow Reinforced High-Strength Concrete (GCHRHSC) short columns, discussing the effects of GFRP tube thickness and reinforcement on this performance. Experimental results reveal that all specimens undergo progressive brittle failure, characterized by concrete crushing and vertical cracking. Increasing the tube thickness significantly enhances the load-bearing capacity, whereas the inclusion of longitudinal reinforcement mitigates the damage extent of the sandwiched concrete. Finite element (FE) models are developed using ABAQUS to evaluate the influence of various parameters-including the winding angle and thickness of the GFRP tube, longitudinal reinforcement diameter, stirrup spacing, as well as the concrete strengths of the Pretensioned spun High-strength Concrete (PHC) pipe pile and the sandwiched concrete-on the axial compressive performance of GCHRHSC columns. As intelligent algorithms, Artificial Neural Networks (ANN) and Genetic Algorithms (GA) are widely applied in engineering fields such as structural performance prediction and design optimization. To meet the extensive data requirements for neural network training, six component parameters are randomly combined to generate a dataset comprising 720 FE models. The total cost of the GCHRHSC column consists of the individual costs of the GFRP tube, the PHC pipe pile, and the sandwiched concrete. The GA designates the ratio of load-bearing capacity to cost (LP) as the optimization objective. During the iterative process, the ANN predicts the load-bearing capacities for various parameter combinations generated by the GA, thereby achieving structural optimization of the GCHRHSC columns by adjusting cross-sectional dimensions and material strengths. Comparisons between the ANN-predicted load-bearing capacities at various convergence points and the corresponding FE results demonstrate that errors are controlled within 10%, validating the accuracy of the GA-ANN algorithm. Consequently, the proposed GA-ANN algorithm is applicable to the structural design and optimization of GFRP confined hollow reinforced high-strength concrete short columns and similar composite columns.