Nvidia Faces a New Threat as Amazon Targets Its Biggest AI Advantage

Amazon AMZN is working to reduce a key barrier to using its artificial intelligence chips by introducing software that lets developers transfer existing code to its processors with minimal changes.

The new TorchNeuron system enables PyTorch-based applications to run on Amazon’s Trainium chips after a single configuration adjustment, according to information presented at a developer conference in Silicon Valley. Amazon plans to make the software available free of charge as an open-source project.

The release puts Amazon in competition with Nvidia NVDA, whose CUDA software has helped developers optimize AI workloads on its graphics processing units. Making applications easier to move between chip platforms could give cloud customers more flexibility when choosing computing infrastructure.

Alphabet GOOGL is also promoting internally developed AI processors as major technology companies seek alternatives to Nvidia’s hardware. Wider support for familiar programming tools may help cloud providers attract developers without requiring them to rebuild existing applications.

Nvidia is expanding its software offering in parallel. The company recently introduced an Open Agent Safety Platform designed to help developers manage AI agents and reduce the risk of models operating outside intended restrictions.

Amazon’s latest move adds to competition over the software ecosystem surrounding AI chips, where developer convenience may influence hardware adoption alongside computing performance and…