In view of the problem that massive spatial data interpolation is a complex and time-consuming computing process, and the traditional CPU implementation methods can't meet the real-time processing demand, in this paper, we propose a parallel bilinear spatial interpolation algorithm, which is accelerated by the graphic processing unit(GPU) and implemented in compute unified device architecture(CUDA). Firstly, we introduce the basic idea of general purpose computing on graphics processing units (GPGPU) and then discuss the technology of the CUDA programming model. Secondly, we introduce the principle of the bilinear interpolation algorithm and analyze the feasibility of mapping the bilinear interpolation algorithm program onto the GPU, and then we provide detail of implementing our parallel bilinear spatial interpolation algorithm on GPU that uses the CUDA programming model. Finally, we conduct several groups of experiments to demonstrate the strength of our GPU implementation method by measuring the performance over standard CPU implementation. The experimental results show that the GPGPU-based parallel algorithm can take full advantage of the GPU's parallel computing capabilities, and can achieve about 40 times speedup; it is able to meet the demand of real-time processing of massive spatial data interpolation.