I think it's less about the size in square feet than the size in dollars (including installation, maintenance, and waste disposal). The pool of people willing to buy a $100M device is small. The pool of people willing to buy a $10M device - even if it's less efficient - is much larger. That makes it a potential revenue source, instead of an eternal revenue sink. Much as fusion doesn't become technically useful until energy output exceeds energy input, it doesn't become economically useful until revenue (for its maker) exceeds cost.