Even their lead engineer thinks that this is possible and talks as if they already had a prototype.... (Tapeout is in August for those who didn't bother to read)
Assuming this is the same engineer that estimated the prior power requirements for their presently shipping product, I'm not sure how anyone can believe anything being reported prior to having a physical working product.
Again, I am using well-documented history as an indicator.
Right, last time the estimate was 0.8 to 1.2watts tops.
Turned out to be anywhere from 1.6 watts to 6 watts per GH/s. [in reality]
This underestimation caused the entire platform to have to be redesigned and caused immense delays of nearly a year!