Argonne flexes spare supercompute to build private AI inference service

Mmmm, what I could do with a cluster of H100s like that. Tim Allen grunt