Much of the data generated by the Internet of things (IoT) remains unanalyzed and is referred to as dark data. Making intelligent IoT applications that benefit society requires efficient analysis of this data. A key component of discovering hidden patterns in these large datasets is the use of artificial intelligence (AI) techniques, particularly clustering algorithms. Although traditional clustering methods like K-means, FM, and FCM are computationally demanding, they are often implemented on high-performance computing (HPC) clusters, which are energy-intensive and face challenges such as communication latency and privacy. These issues can be addressed by using edge computing to enable lightweight computations at the network edge. An analysis of the performance and power consumption of edge computing architectures equipped with low-power GPUs is presented in this paper, which focuses on the NVIDIA AGX Xavier platform. By comparing edge platforms with their high-performance cloud counterparts, we evaluate the efficiency of clustering algorithms implemented on each. In comparison with sequential clustering, GPU-accelerated clustering achieves up to 11× performance gains and 150% energy savings compared to HPC-based clustering. IoT applications that utilize AI can benefit from edge computing’s energy efficiency and scalability.