Feature size on a chip scales via square (i.e., you get 4x more transistors on a 45 nm chip than on a 90 nm chip, all else being equal) So its more like 100-400x as dense, not 10-20x.

However, I assume they are talking about the minimum size it could scale to, not the size of the first generation product. Either way, I'm sure a lot of development would be required to mass produce it. IBM's MRAM has been 'a few years away' for years now, it is starting to remind me of bubble memory, the next big thing back in the 80s. Hopefully this technology will not suffer the same fate, but time will tell.

... that if I store all ones (or all zeroes) I can make my memory more valuable by converting it into a diamond?

Was bursting by late 1970s. It had promise in the early 1970s.

It's serial with major and minor loops, so it's doubtful it will even make a comeback. Still, it beat anything else at the time for robust nv storage. More like low capacity solid state drive for Military.

Yeah, think so. Not the most fun way of going about it though....

http://news.bbc.co.uk/1/hi/world/americas/7725815.stm

They use the obvious-when-you-think-about-it pulsed liquid injection chemical vapor deposition technique. I've been drinking a lot of tequila recently resulting in quite a bit of pulsed liquid injection and plenty of chemical vapour. Doctor says my crystals are uric not diamond, though.