Google and Amazon challenge Nvidia's AI dominance with dedicated PyTorch tools

TorchTPU and TorchNeuron aim to allow AI apps to run on alternative hardware without rewriting code

By LineZotpaper

Published20 October 2026

Read Time2 min

Sources2 outlets

Google has released TorchTPU, an open-source version of the PyTorch framework optimized for its Tensor Processing Units (TPUs). The announcement comes a day after Amazon released a similar tool, TorchNeuron, for its Trainium chips. Both moves represent a direct effort by the largest cloud providers to erode Nvidia's dominant position in the AI chip market, where most models are built and run using Nvidia's GPUs.

Google announced on Tuesday the general availability of TorchTPU, its adaptation of the widely used PyTorch machine learning framework for Google's own TPU hardware. The release follows Amazon's rollout of TorchNeuron on Monday, which targets Amazon's Trainium chips.

PyTorch, originally created by Meta and now under the stewardship of the Linux Foundation, is the dominant toolkit for building and training AI models. According to some accounts, it is used by roughly 85 percent of AI applications. Because Nvidia's GPUs were the standard hardware for AI research when PyTorch was developed, much of the toolkit was built to work specifically with Nvidia's hardware ecosystem.

Both Amazon and Google are working to change this. Amazon stated it is going a step further by submitting code directly to the main PyTorch project to decouple its internal features from Nvidia's chips, making the framework more hardware-agnostic by default.

Google has focused on ecosystem partnerships, stating it worked directly with Meta, Hugging Face, and other AI development firms to ensure TorchTPU works well with popular tools. Google also launched a dedicated 'TPU agent suite' to help developers migrate applications to its cloud infrastructure and troubleshoot errors.

While these tools do not guarantee developers will switch from Nvidia hardware, they lower the technical barriers to using alternative chips, potentially reshaping the competitive landscape of the AI industry.

Analysis

Why This Matters

The tools directly challenge Nvidia's 'software moat,' the deep integration of its CUDA platform with AI frameworks like PyTorch.
They strengthen the value proposition of Google Cloud and AWS for AI workloads, potentially driving more cloud revenue to these platforms.
If widely adopted, these tools could lead to greater competition in AI hardware, potentially lowering costs for developers and startups.

Background

Nvidia's GPUs have been the dominant hardware for AI training and inference for over a decade. This dominance is partly due to the CUDA software platform, which became the standard way to program AI models. Most popular AI frameworks, including PyTorch, were built with CUDA as a primary target, making it difficult for competing chips to gain traction. Major cloud providers like Google (with TPUs), Amazon (with Trainium), and Microsoft have invested heavily in developing their own AI chips, but adoption has been limited by software compatibility.

Key Perspectives

Google and Amazon: They argue for a more open and competitive hardware ecosystem. By releasing their own optimized versions of PyTorch, they aim to remove software compatibility as a barrier to entry, giving developers a genuine choice of hardware for AI tasks. Nvidia: The company remains the market leader with a deeply entrenched ecosystem. While these tools target its core advantage, Nvidia continues to develop both its hardware (GPUs) and software stack to maintain its lead. AI Developers and Startups: They stand to benefit from increased competition. The ability to easily switch between hardware platforms offers potential cost savings and better availability.

What to Watch

Adoption benchmarks and performance comparisons of TorchTPU and TorchNeuron against native Nvidia PyTorch.
Whether Nvidia responds with software updates or partnerships to defend its ecosystem.
The reaction of the open-source community to Amazon's proposed changes to the upstream PyTorch project.

Sources

Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers · TechCrunch
Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers · AI News & Artificial Intelligence | TechCrunch

Checked

7 claims in this story checked against 1 source: 2 supported.

Show each claim

Partly supportedGoogle and Amazon are challenging Nvidia’s AI dominance with dedicated PyTorch tools, and are described as the largest cloud providers.TechCrunch: “Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers.”
SupportedTorchTPU and TorchNeuron aim to let AI apps run on alternative hardware without rewriting code.TechCrunch: “Developers can do this with a simple configuration change and no re-writing of their apps. Amazon’s own version of PyTorch, called TorchNeuron. TorchNeuron allows AI apps to be ported to Amazon’s homegrown AI chip, Trainium, and to AWS with no rewriting.”
Partly supportedGoogle released TorchTPU as an open-source PyTorch version optimized for its TPUs.TechCrunch: “Google announced that TorchTPU, its long-awaited version of PyTorch, is now available as an open-source project. ... Google’s version, TorchTPU, lets AI apps run on Google’s AI chips, called Tensor Processing Units (TPUs).”
Partly supportedGoogle announced TorchTPU’s general availability on Tuesday, a day after Amazon released TorchNeuron.TechCrunch: “On Tuesday, Google announced that TorchTPU, its long-awaited version of PyTorch, is now available as an open-source project. ... Amazon made a similar announcement Monday, when it released its own version of PyTorch, called TorchNeuron.”
Partly supportedMost models are built and run using Nvidia GPUs.TechCrunch: “most AI apps are written by default to run on Nvidia’s chips, or GPUs.”
SupportedPyTorch is used by roughly 85 percent of AI applications, according to some accounts.TechCrunch: “It is, by some accounts, used by 85% of the AI apps.”
Partly supportedAmazon is submitting code to the main PyTorch project to decouple internal features from Nvidia chips and make the framework more hardware-agnostic by default.TechCrunch: “It is submitting code to the main PyTorch project that decouples various features from Nvidia’s chips, making it more open to any hardware by default.”