One of the things that the fusion community is trying to do is scale down the size of machine that is required, basically the view is that the volume of facility determines the cost (interestingly the stats presented seem to support this) and by building compact devices the price of the machine will decline accordingly. This will enable commercial implementations - if it works!