# Google and Amazon take aim at Nvidia’s hold on AI developers

URL: https://ai-pulse.aglaganova.com/blog/google-and-amazon-take-aim-at-nvidias-hold-on-ai-developers
Published: 2026-10-22
Author: Roman Aglaganov, Full-stack developer
Category: Industry
Region: Global
Tags: chips, coding

> Most AI software is written with PyTorch, which was built around Nvidia’s chips. This week Amazon and Google released versions that run on their own chips with a simple configuration change.

![Illustration: a thick black line splitting into three paths towards three grey squares, with one path in red](https://ai-pulse.aglaganova.com/api/media/file/torch-chips-1.png)

Nvidia’s lead in AI is not only about hardware. Most AI apps are built with PyTorch, the open-source toolkit for training and running models, and by some accounts 85% of AI apps use it. When Meta created PyTorch in 2016, Nvidia’s graphics chips were already the standard for AI research, so many of its features were built around them. This week Amazon and Google both moved to change that.

## What they released

**Amazon, Monday:** TorchNeuron, a version of PyTorch that runs on Amazon’s own Trainium chips, so apps can move to them and to AWS without rewriting. Amazon says it is also submitting code to the main PyTorch project that loosens features from Nvidia’s chips, so PyTorch works better with any hardware by default.

**Google, Tuesday:** TorchTPU, its long-awaited version of PyTorch for its Tensor Processing Units, released as open source at the PyTorch Conference. Google says existing PyTorch apps can move to TPUs, and to Google Cloud, with a configuration change. It worked with Meta, Hugging Face and the makers of the popular serving tools vLLM and SGLang to make sure they work with it, and released an agent suite to help developers move their apps and fix errors.

## Will developers move?

Not automatically, TechCrunch noted. Nvidia’s software and the habits of a generation of AI engineers are deeply rooted, and an easier move does not guarantee a cheaper or faster one. But the cost of trying another chip has dropped.

## Why it matters

AI companies are spending hundreds of billions of dollars on computing, and Nvidia takes a large share of it. Every app that can run unchanged on other chips gives buyers more bargaining power, and makes Google’s and Amazon’s own chips a more realistic alternative.

## Sources

- [TechCrunch: Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers](https://techcrunch.com/2026/10/20/google-has-joined-amazon-in-attacking-nvidias-stronghold-with-ai-app-developers/)
- [PyTorch](https://pytorch.org/)