Google Cloud launches TPU 8t and 8i chips with 3x faster training, challenging NVIDIA's AI infrastructure dominance

Google Cloud announced eighth-generation TPUs: TPU 8t for model training (3x compute performance, 10x faster storage, 2x data transfer versus predecessor) and TPU 8i for inference, with the ability to cluster over 1 million TPUs together for 80% better performance-per-dollar. Google CEO Sundar Pichai disclosed at the same event that 75% of new Google code is now AI-generated. The announcement also included a deepened ServiceNow partnership to deploy autonomous enterprise agents.