Google and Amazon release PyTorch tools to rival Nvidia AI chips
Tech giants release new open-source tools allowing AI developers to move their applications from Nvidia hardware to custom chips from Google and Amazon.
Google has released TorchTPU to help developers run PyTorch applications on its custom Tensor Processing Units.
Amazon launched TorchNeuron, allowing AI models to migrate to its Trainium chips with minimal code changes.
Both companies aim to break Nvidia's dominance, as roughly 85% of AI apps currently use the PyTorch framework.
Amazon is also contributing code to the main PyTorch project to make the software less dependent on Nvidia hardware.