Quote from the PopPhoto article linked at the bottom of the post:
The shot is composed of 10,000 individual shots, taken over 6 hours spread over two days, from the 20th floor of the Tokyo Tower last September. Taken from four points on a viewing platform on the building, photographer Jeffrey Martin used a Canon 7D with a 400mm f/5.6 lens. In order to take all the photos, they used the Clauss Rodeon VR Head ST gigapixel rig.
It took an incredible 12 weeks of computer time to crunch all the images into one file, even when using a powerful rig. Fujitsu provided Martin with the processing power to put the whole thing together, using a workstation with dual 6-core CPUs, 192GB of RAM, and a 4GB graphics card.