Google Launches TPU 8t & 8i to Challenge Nvidia in AI Chip Race

Google Cloud launched eighth-gen TPUs—TPU 8t for training and TPU 8i for inference—at Google Cloud Next 2026, claiming 3x faster training, ~80% better performance-per-dollar, and scalability to over one million chips for exaflops-scale AI workloads. Initial coverage highlighted the reveal as a boost to Google's AI infrastructure; latest details position it as a cost-competitive alternative to Nvidia GPUs while offering complementary Nvidia Vera Rubin support. The development strengthens Alphabet's cloud/AI edge, potentially driving GOOG revenue and margins if adoption surges, though Nvidia's ecosystem remains a barrier.