Borehole acoustic measurements are essential in the exploration and development of resources such as oil, gas, and geothermal energy. Three-dimensional numerical simulation allows for the analysis of how complex borehole environments influence the acoustic wave propagation, providing a foundation for accurate inversion of elastic properties of formation and cement annulus. However, the current finite difference methods for simulating elastic waves often require significant computational resources, especially when handling large-scale three-dimensional models, leading to substantial increases in both computation time and storage requirements. This paper proposes a multi graphics processing unit (GPU) parallel numerical simulation algorithm based on CUDA (Compute Unified Device Architecture) for simulating the acoustic field in a 3D borehole environment. Using the simulation of ultrasonic guided waves in a cased well as an example, the algorithm efficiently divides the model, allocates resources, and coordinates computations to leverage the computational power of multiple GPUs. The results demonstrate that the algorithm achieves a speedup of 8.93 for small-scale models on a single GPU and a speedup of up to 9.95 for large-scale models when using four GPUs. The multi-GPU parallel algorithm proposed in this study provides an efficient solution for the numerical simulation of complex 3D borehole wave fields and offers new insights for acoustic logging simulations based on high-performance GPU parallel computing.