Google Open-Sources TorchTPU to Take On Nvidia
Google open-sourced TorchTPU, letting PyTorch apps run on its TPUs with a simple config change. It joins Amazon in challenging Nvidia AI hardware dominance.
Google has open-sourced TorchTPU, its version of the popular PyTorch AI toolkit that runs on Google's own Tensor Processing Units. The move puts Google right behind Amazon in a direct challenge to Nvidia, whose chips power most AI apps built today.
What TorchTPU does
PyTorch is the open-source toolkit AI developers use to build, train, and run AI models. Some estimates say about 85 percent of AI apps use it. It was built by Meta back in 2016, when Nvidia GPUs were already the standard, so many of its features were designed around Nvidia chips. Meta handed it to the Linux Foundation in 2022. Google's TorchTPU changes that default. It lets existing PyTorch apps run on Google's Tensor Processing Units, and it makes moving an app to TPUs - and therefore to Google Cloud - as simple as a configuration change, with no rewriting of the code. Google announced the release at PyTorch Conference 2026 with the code available on GitHub.
Google joins Amazon's attack on Nvidia's grip
Amazon made a similar move one day earlier, releasing TorchNeuron, its own version of PyTorch for Amazon's Trainium chips on AWS. Amazon is going further: it is submitting code to the main PyTorch project that separates various features from Nvidia's chips, so PyTorch is open to any hardware by default. Google may be one-upping Amazon in a few ways. It worked directly with Meta, Hugging Face, Inferact (the maker of vLLM), and RadixArk (the maker of SGLang) to make sure TorchTPU's features work with the tools developers already use, just like the original PyTorch. Frameworks such as vLLM, SGLang, TorchTitan, VeRL, and Ray already support it, and in Hugging Face Transformers v5, models can be deployed on TPUs by setting device_map to "tpu". Google also launched a TPU agent suite that helps developers move their apps to Google's chips and cloud while fixing errors along the way. It even updated its LiteRT tool, which runs AI on phones, laptops, and embedded devices, to work better with PyTorch apps.
What this means for developers
None of this means developers will drop Nvidia's hardware overnight just because alternatives are easier to reach. But two of the biggest cloud providers are now clearly working to end Nvidia's advantage. For developers, the practical win is choice: PyTorch code written for Nvidia GPUs can now move to TPUs or Trainium with minor changes, and the tooling to make that move is open source.
Sources
TechCrunch - https://techcrunch.com/2026/10/20/google-has-joined-amazon-in-attacking-nvidias-stronghold-with-ai-app-developers/ NewsBytes - https://www.newsbytesapp.com/news/science/google-takes-on-nvidia-with-new-ai-developer-tool/story PANews via Binance - https://www.binance.com/en/square/post/10-21-2026-ai-trends-google-open-sources-torchtpu-to-challenge-nvidia-s-ai-ecosystem-dominance-379518544710037