Amazon Releases Native PyTorch Backend for Its Custom AI Chips, Challenging Nvidia's Software Dominance

Amazon unveiled TorchNeuron, a native PyTorch backend for its custom Trainium and Inferentia chips, marking a direct challenge to Nvidia's deep integration in AI framework development. The technology, released in public beta on October 20, 2026, allows developers to run existing PyTorch code on AWS-built accelerators with minimal code changes—typically just a single configuration line—and is available as open-source software under the Apache 2.0 license.