Skip to content

All the news that matters, in plain English.

Latest
  1. Foden hat-trick as City rout AEK 5-1; Liverpool edge Villarreal 4-3
  2. PSG beat Barcelona 4-2 in Champions League thriller at Parc des Princes
  3. AEP, Amazon and X-energy explore 960 MW advanced nuclear project in Indiana
  4. OpenAI acquires Orchestra Bio to push AI agents into drug development
  5. SpaceX Q3 revenue jumps 161% to $14.58bn as AI and Starlink beat forecasts
  6. Netflix beats profit forecasts but slides on soft fourth-quarter outlook
  7. Anthropic in talks with Nscale over second $45bn AI data centre deal
  8. HPE's Spaceborne Computer-4 to run first AI workloads and HPC benchmarks on the Moon

Technology

Google open-sources TorchTPU, letting PyTorch apps run on its TPUs with a config change

Google has released TorchTPU as open source, a native PyTorch stack that lets AI apps written for Nvidia GPUs run on Google's TPUs — and therefore Google Cloud — with a simple configuration change and no rewrite.

By James Calloway · · 1 min read

AI-generated editorial illustration of a glowing processor chip bearing the Google G, rising over a stylised cloud datacentre landscape — an illustration for the story on Google open-sourcing TorchTPU.
Illustration: Flip News (AI-generated)

Google has released TorchTPU, its long-awaited native PyTorch stack for its Tensor Processing Units, as an open-source project on GitHub — joining Amazon in a direct challenge to Nvidia’s grip on AI development (TechCrunch).

Announced on Tuesday at PyTorch Conference 2026, TorchTPU lets AI apps built with the popular PyTorch framework run on Google’s TPUs, and therefore on Google Cloud, with a simple configuration change and no rewriting of their apps (Google Cloud). The stack was co-developed with the PyTorch team at Meta, and Google says it worked directly with Meta, Hugging Face, Inferact (the maker of vLLM) and RadixArk (the maker of SGLang) so TorchTPU’s features work with the tools developers already use (TechCrunch; Google Cloud).

The move mirrors Amazon’s announcement a day earlier of TorchNeuron for its Trainium chips — also with no rewriting required — with Amazon going further by submitting code to the main PyTorch project that decouples its features from Nvidia’s chips (TechCrunch).

Google also launched a dedicated TPU agent suite, now in preview on GitHub, to help developers move their apps to its chips and cloud and fix errors along the way, and updated its LiteRT on-device AI tool to work better with PyTorch AI apps (TechCrunch; Google Cloud).

The framing from TechCrunch is blunt: most AI apps are written by default to run on Nvidia’s GPUs, and Amazon and Google are now “clearly working to end Nvidia’s advantage” — though neither expects developers to ditch Nvidia hardware en masse overnight (TechCrunch).

More on this topic: all Technology stories

Get Flip News by email

This opens your email app — we add you manually. No account, no spam, no third parties.