But not at the single atom level, tunneling becomes a much much bigger problem -- it's definitely exponential.
the bbc article doesn't say, but my guess is that it has to be super cooled to work. Which makes a lot of sense actually considering that this group says that on the bigger picture they're working on Quantum Computing, which generally is only done in the laboratory at temperatures pushing our ability to cool things down near abs zero.
So these guys probably don't care that to actually use this set of atoms as a transistor, you may have to be really cold, or even use a really low clock cycle, because their bigger goal is to create a quantum computer, which would be soo powerful for certain things that having to query it at a very low frequency, and have it be in the lab are OK things
I wonder if, as gradual progress is made on these issues, some of the techniques can be back-scaled to larger die sizes for multi-layered chips. This happens to some extent already - current Intel and AMD chips have 10 layers or more.
IBM made a minor splash (groan) a few years ago when they unveiled water-cooled chips that had liquid flowing though 50u channels. They're in partnership with a Swiss university on a technology called Aquasar (http://www.youtube.com/watch?v=FbGyAXsLzIc), the first incarnation of which is anticipated to go online this year.
