On Tuesday Google announced that TorchTPU, its long‑awaited version of PyTorch, is now released as an open‑source project. This move puts Google alongside Amazon in an effort to weaken Nvidia’s grip on the AI market, where most AI applications are built to run on Nvidia’s GPUs by default. PyTorch, the widely used open‑source framework for building, training and deploying AI models, is said to power about 85 % of AI applications. It was created by Meta (formerly Facebook) in 2016, well before large language models reshaped the industry. At that time Nvidia’s GPUs were already the standard hardware for AI research, so many PyTorch features were designed specifically for them. In 2022 Meta transferred stewardship of PyTorch to the Linux Foundation, a nonprofit that oversees many key open‑source projects.
TorchTPU enables AI workloads to run on Google’s Tensor Processing Units (TPUs) and simplifies the migration of existing PyTorch applications to TPUs and Google Cloud. Developers can make the switch with a simple configuration change, without rewriting code. Amazon made a comparable announcement on Monday, unveiling TorchNeuron, its own PyTorch variant that lets developers port models to Amazon’s Trainium chip and run them on AWS without code changes. Amazon also said it is contributing code to the main PyTorch repository that separates various features from Nvidia‑specific dependencies, making the framework more hardware‑agnostic by default.
Google may have an edge in a few areas. The company says it worked directly with Meta, Hugging Face, Inferact (the creator of vLLM), RadixArk (the maker of SGLang) and other partners to ensure TorchTPU’s features and libraries integrate smoothly with these popular AI tools, just as the original PyTorch does. Google also launched the TPU Agent Suite, a set of utilities to help developers move their applications to TPUs and Google Cloud and to troubleshoot issues. In addition, the firm updated its LiteRT tool, which runs AI models on phones, laptops and embedded devices, to improve compatibility with PyTorch‑based applications.
These developments do not guarantee that AI developers will abandon Nvidia hardware en masse, but both Amazon and Google are clearly working to erode Nvidia’s advantage in the ecosystem.
Julie Bort, Venture Editor, TechCrunch.