The interesting thing for me is that there is a real overlap between the locations of datacentres. It's obvious why this happens, but the possibilities it opens that's really cool. If you include the backhaul each provider has all the clouds suddenly get very close.
We thought about doing that, it's easier to do for North America because the trial of clues is easier to follow for an English speaker (e.g. DCK article yesterday about the new DC in Oregon). However, we'd like to do it globally too.
I'm going to predict a lot of the DCs will again fall next to fat pipes off-continent.
You should be able to get the data programmatically, most isps have looking glass servers that you can look at to figure out who is peering with who. You can also look their AS #s up on Robtex and see if they're peering..
