Google unveils TorchTPU in a push to break Nvidia’s stranglehold on AI compute

Google has launched TorchTPU, a version of the PyTorch AI framework optimized for its Tensor Processing Units (TPUs), aiming to reduce Nvidia's dominance in AI computing by easing model deployment on TPUs without code rewrites. This move follows AWS's introduction of TorchNeuron for its Trainum chips, both efforts to diversify AI hardware compatibility beyond Nvidia GPUs.