Well said. The only "counter" I could give you is that another lever in token prices is speed, which requires very expensive hardware that most providers don't have access to (i.e., NVIDIA). As well as the fac tthat those tokens run in low batches which increase the cost per token, so they are inherently more expensive.

But I do contend as much as you about token prices not cliff diving more and more as time goes by. Hopefully, hardware improvements will drop cost per tokens faster (which has been the case for now, but might not be the case in the future).