Because you can't stack that much ram on a GPU without sufficient channels to do so. You could probably do 64GB on GDDR6 but you can't do 128GB on GDDR6 without more memory channels. 2GB per chip per channel is the current limit for GDDR6 this is why HBM was invented.

It is why you can only see GPUs with 24GB of memory at the moment.

HBM2 can handle 64GB ( 4 x 8GB Stack ) ( Total capacity 128GB )

HBM3 can handle 192GB ( 4 x 24GB Stack ) ( Total capacity 384GB )

You can not do this with GDDR6.