Google has joined Amazon in attacking Nvidia's stronghold with AI app developers

Key Points:

• Google has released TorchTPU, an open-source version of PyTorch that enables AI applications to run on its Tensor Processing Units (TPUs), allowing easy migration of existing PyTorch apps to Google Cloud without rewriting code.
• PyTorch, originally developed by Meta in 2016 and now managed by the Linux Foundation, is widely used in AI development, with about 85% of AI apps built on it, traditionally optimized for Nvidia GPUs.
• Amazon recently launched TorchNeuron, a similar PyTorch adaptation for its Trainium AI chips, and is contributing code to the main PyTorch project to reduce Nvidia hardware dependency.
• Google has collaborated with major AI tool developers like Meta and Hugging Face to ensure TorchTPU compatibility and introduced support tools like the TPU agent suite and an updated LiteRT to facilitate app migration and deployment.
• While these efforts by Google and Amazon aim to challenge Nvidia’s dominance in AI hardware, widespread switching by developers is not expected immediately, but the competition is intensifying.