Google announced TorchTPU, its long-awaited version of PyTorch, is now available as an open source project
Rhea Lease · AI writer ·
0 primary · 2 independent · 3 commentary · 0 vendorImage courtesy of Fay Tograph · AI photographer, made with gpt-image-2
TechCrunch reported that Google announced TorchTPU, its long-awaited version of PyTorch, is now available as an open source project.[1]The release lets developers run existing PyTorch models on TPUs with minimal code changes, using familiar APIs, according to Google Cloud.[2]
Google LLC is striving to pry open Nvidia Corp.'s grip on the artificial intelligence computing industry with the launch of a specialized version of the open-source PyTorch framework, called TorchTPU, which is designed to eliminate a key advantage of its rival.[3]
Google did not submit such code but said it collaborated closely with Meta, Hugging Face, Inferact and RadixArk to ensure that various TorchTPU features and code libraries are compatible with their developer tools, SiliconANGLE reported.[3][1]The company also launched the TPU agent suite to help migrate applications to Google Cloud and updated LiteRT, which Google said should now be compatible with PyTorch-based apps.[3]
TorchTPU is now available on GitHub as an open-source project, Google Cloud said, allowing developers to train, serve and deploy models on TPUs using standard PyTorch workflows.[2]
What people on X are saying, summarized by Grok
On X, TorchTPU is widely cast as pressure on Nvidia's software lock-in, often paired with Amazon's similar PyTorch release a day earlier. Many say the real barrier was rewriting code, not the chips, and some cite Meta already running related TPU inference in production as a sign big customers gain leverage. Skeptics say a model that runs is not one that matches CUDA speed, pointing to missing features, extra tuning, and price claims that still need checking on real workloads.