Skip to main content

Kaer NewsSoftware

Amazon launches TorchNeuron so PyTorch apps run on Trainium chips

Amazon Web Services released TorchNeuron on 19 October, letting PyTorch apps run on its Trainium chips with one configuration change, entering public beta on 20 October 2026.

3 min read
Amazon launches TorchNeuron so PyTorch apps run on Trainium chipsPhoto: TechCrunch

Key points

Amazon Web Services released TorchNeuron, an open-source tool letting PyTorch apps run on its Trainium chips with one configuration change, entering public beta on 20 October 2026.

Amazon Web Services released TorchNeuron on 19 October, an open-source tool that lets applications written with PyTorch run on Amazon's Trainium AI chips by changing one line of configuration. The tool enters public beta on 20 October 2026 under the Apache 2.0 licence and supports Trainium2 and Trainium3 chips. Amazon says no code rewrite is required.

The release targets Nvidia's software advantage. Amazon says about 85% of AI apps are written with PyTorch, the free open-source toolkit developers use to build, train and run AI models. Facebook originally developed PyTorch, and it has been part of the Linux Foundation since 2022. Nvidia has spent years ensuring its CUDA software — the infrastructure that lets apps run on Nvidia GPUs — seamlessly supports PyTorch apps, embedding its hardware in the toolkit by default.

What the tool actually does

TorchNeuron is adapted from PyTorch, so developers who know PyTorch already know how to use it. It works by changing one line of configuration, according to Amazon. Some PyTorch features are not yet supported. Amazon is not the first company to bring its chips to the platform: PyTorch users can also port apps to AMD GPUs or Apple's chips with a simple configuration change.

Amazon says more than 30 early customers in robotics, voice AI and world models have tested TorchNeuron. The tool adds to Amazon's broader Neuron software effort, which includes tools for programming and optimising workloads on its custom chips. Google and Microsoft are also pursuing ways to shift AI workloads toward their own processors, intensifying industry efforts to reduce reliance on Nvidia.

Colin Brace, Vice President of Technical Software at AWS Annapurna Labs, the unit that makes Amazon's custom chips, argued that porting to Trainium gives startups access to Amazon's entire cloud rather than a single-purpose provider. "You're building on AWS. You're not betting on a single-purpose compute provider and scrambling to build everything else around it," Brace said via email.

How Amazon argues the economics

AWS says its cloud can save startups money because the same chip handles both training — teaching the AI model — and inference — putting the finished model to work — and is tightly integrated with cloud infrastructure, even though AWS does not offer the cheapest per-token price. Brace cited Deepgram, an AI voice app provider, which moved to Trainium using native PyTorch support and now processes 70% more audio per dollar at production-grade accuracy.

Amazon is also pushing PyTorch to be less automatically centred on Nvidia and will contribute the code to do that to the main PyTorch project. In the blog post announcing the tool, Amazon wrote that it is working with the PyTorch community to remove code that ties the framework to specific chips, with the goal of making PyTorch more hardware-agnostic for the entire developer community.

The move is not without uncertainty. One of PyTorch's purposes is to hide the difficulties of getting AI code to run on hardware, and it remains unclear whether pushing Nvidia out further gives developers more choice and more affordable options, as Amazon promises, or simply more configuration headaches. AWS frames the change as removing lock-in. "The real story is removing the lock-in that kept developers from having a real choice at all," Brace said.

Frequently asked questions

What is TorchNeuron?

TorchNeuron is an open-source tool released by Amazon Web Services that lets applications written with PyTorch run on Amazon's Trainium AI chips by changing one line of configuration, with no code rewrite required. It supports Trainium2 and Trainium3 under the Apache 2.0 licence.

When does TorchNeuron enter public beta?

TorchNeuron enters public beta on 20 October under the Apache 2.0 licence, supporting Trainium2 and Trainium3 chips. Amazon Web Services released the tool on 19 October.

How many customers have tested TorchNeuron?

Amazon says more than 30 early customers in areas including robotics, voice AI and world models have tested TorchNeuron. AWS also cited Deepgram, which processes 70% more audio per dollar after moving to Trainium.

More from this edition