I don't get it. There are plenty of fully functional open weight models that run on regular consumer hardware (GPU or unified memory architecture).
We're way past the point where we know this technology is going to be with us even after the current VC-fueled overspending bubble has burst.
