Next-gen nuclear fusion alloy discovery reduced from months to hours with new US tool

Ames National Laboratory expanded DuctGPT, an AI-assisted materials design tool, to predict alloy compositions for fusion reactors.

The system fine-tunes NIST’s AtomGPT with materials data and can suggest candidate elements quickly on a desktop computer.

It is helping identify tungsten-based alloys that preserve high melting points while improving ductility.

The approach could cut the time, cost, and compute needed to find materials that can withstand extreme heat and radiation in fusion reactors.