Could you imagine getting $100m dollars to replicate an #OpenSource model?
Look, I know all too well there's training data to collect, try to filter out the garbage (if getting it from untrusted sources), and training to do, but still... I couldn't even imagine how to spend that quantity of money on a project like this
It doesn't take that many people, and at some point it doesn't make sense to keep piling on more hardware.