PyTorch Conference 2026: Google and Meta Open-Source TorchTPU for Native PyTorch on TPUs
Google and Meta open-sourced TorchTPU, a native PyTorch backend for Google's TPUs, at PyTorch Conference North America 2026, and Meta said it plans to move larger LLM workloads onto TPUs with it. AWS released TorchNeuron for Trainium the same week, while developers questioned TPU availability and where vendor lock-in now sits. By Andrew Hoblitzell
Google and Meta have jointly open-sourced TorchTPU, a native PyTorch backend for Google's Tensor Processing Units (TPUs). TorchTPU aims to simplify the process of porting models to TPUs by allowing a simple change from model.to(cuda) to model.to(tpu). This development comes as TPUs join GPUs and CPUs as a native option for PyTorch, with TPUs handling larger-scale LLM workloads.
The open-source TorchTPU joins four other repositories, including vLLM-TorchTPU, SGLang-TorchTPU, and TorchTitan, and has been tested for near-linear scaling with up to 8,000 TPU chips. Google's TPU agent suite is also now in preview, automating onboarding, diagnostics, and performance tuning for practitioners.
Brief written by urgent.news from InfoQ's own syndicated text. Machine-written — may contain errors; check the original before relying on it.
