Key facts
- TorchNeuron allows PyTorch code to run on Amazon's Trainium AI chips with a single configuration change.
- PyTorch is the most popular AI coding language, stewarded by Meta.
- Nvidia has a long-standing software tool called Cuda that optimizes PyTorch code for its chips.
- Amazon's move aims to reduce the number of steps for AI developers to try its in-house chips.
- Early customers report steep cost savings when switching to Amazon's chips.
Amazon.com is introducing TorchNeuron, a new tool designed to allow its custom AI chips to run code written in PyTorch, the most popular AI coding language, without modification. This move by Amazon's Annapurna Labs chip unit aims to challenge Nvidia's long-standing dominance in the AI computing market, which is underpinned by its Cuda software. PyTorch is stewarded by Meta, and other cloud providers like Google are also developing better PyTorch tools to attract developers. Colin Brace, vice president of software for Annapurna Labs, stated that the goal is to meet customers where they are, reducing the steps required for AI developers to adopt Amazon's in-house chips. Over 30 early customers in fields like robotics, world models, and voice AI are already using the software and reporting significant cost savings.