GCP Cloud Digital Leader Practice Question

Cloud TPU offers Google-designed Tensor Processing Units that are purpose-built to accelerate TensorFlow and other machine-learning workloads. Because the hardware is exposed as a managed service, teams can obtain massive parallel compute power without configuring or scaling clusters themselves. Compute Engine custom machine types with GPUs provide general-purpose GPU access but require more infrastructure management. Vertex AI Feature Store manages features, not training hardware, and BigQuery ML lets users build models with SQL but does not expose specialized training chips. Therefore, Cloud TPU is the most appropriate choice.