Amazon launches TorchNeuron to let AI developers run PyTorch code on Trainium chips

Amazon wants to make swapping out Nvidia chips as simple as changing a configuration setting. The company's new TorchNeuron software backend lets developers move existing PyTorch code onto Amazon's Trainium AI chips by changing one line, switching from .to('cuda') to .to('neuron'), with no rewriting or revalidation required.

Amazon's Annapurna Labs unit announced TorchNeuron on October 19, 2026, with the public beta launching the next day at the PyTorch Conference. This timing was strategic, as PyTorch is the dominant framework for AI model development, allowing Amazon to directly challenge Nvidia in its core territory.

How TorchNeuron works

Released under the Apache 2.0 open-source license, TorchNeuron functions as a backend that integrates directly with PyTorch. It allows Trainium chips, specifically the Trainium2 and Trainium3 generations, to appear to the framework as native devices. The practical effect is that teams running workloads on Nvidia hardware could migrate to AWS Trainium infrastructure without touching their existing codebase.