Google Cloud launches two new AI chips to compete with Nvidia

Google Cloud announced its eighth-generation TPUs split into two specialized chips: the TPU 8t for model training and TPU 8i for inference. The new chips promise up to 3x faster training, 80% better performance per dollar, and clusters of 1 million+ TPUs. Despite this, Google is not replacing its use of existing and upcoming (including the new 'vera rubin') chips from nvidia in its cloud. Google and nvidia are also collaborating to improve the efficiency of nvidia-based systems in Google's cloud via the open-source networking technology called falcon. The broader narrative is that hyperscalers building custom AI chips may reduce nvidia dependence over time, but nvidia's dominance remains intact for now.