You guys might want to take a look at this EETimes article from three years ago, before taking this announcement.
Indeed, another example of how duct tape remains the ultimate engineering tool. Comments about the need for more information to understand this are certainly understandable. The nitty gritty details remain sketchy at this point. Presumably we will find out at least a little bit more once IBM and 3M make the announcement (which the WSJ report indicates will happen this week).
Certainly, from server blade experience, we need to flow air or other fluid in between chip planes because stagnant medium will warm up which is inadequate for heat sinking. So I had thought they were heading in some microfluidic direction.