This paper presents an effective scheme for clustering a huge data set using a commodity programmable graphics processing unit(GPU). Due to GPU’s application-specific architecture, one of the current research issues is how to bind the rendering pipeline with the data-clustering process. By taking advantage of GPU’s parallel processing capability, our implementation scheme is devised to exploit the multi-grain single-instruction multiple-data (SIMD) parallelism of the nearest neighbor search, which is the most computationally-intensive part of the data-clustering process. The performance of our scheme is discussed in comparison with that of the implementation entirely running on CPU. Experimental results clearly show that the parallelism of the nearest neighbor search allows our scheme to efficiently execute the data-clustering process. Although data-transfer from GPU to CPU is generally costly, acceleration by GPU is significant to save the total execution time of data-clustering.