Technology ❯ Hardware ❯ Chips ❯ TPUs
The push targets the CUDA lock-in that keeps PyTorch workloads on Nvidia GPUs.