Google's Tensor Processing Unit (TPU) is a family of application-specific integrated circuits built for machine learning workloads. TPU generations range from the original TPU v1 through TPU v4 and v5p, up to the Trillium generation and the announced Ironwood systems. TPU Pods combine multiple chips with a high-speed interconnect into larger training and inference clusters.

Google Cloud delivers TPUs through its cloud infrastructure as TPU VMs and multislice clusters. The system-of-pods architecture lets customers scale workloads across thousands of chips, with Ironwood positioned as the next-generation platform for large-scale AI training and inference serving.