So here Apple has basically said, screw it, let's focus on the long tail of developers and leave the huge foundation model development to Nvidia. But that was really how it was always going to play out given the big Cuda dependency most large model training has.

→ View original post on X — @mattlynley