This is a bit misleading since Triton is a bit higher level than CUDA. But the idea is kind of right - there’s active development of AMD and Intel backends, and Pytorch is investing into Triton as well.