Ahhhh... I see the problem, here. You assume that for all 7 hours of sunlight, the power output will be at the maximum possible. This is false, of course. The workaround is to use insolation times area times external quantum efficiency.
no, I'm not making that assumption. I am at a similar latitude as he is, and that is based on a daily average. Now, he may have less, because of cloudy days, but we average about 7 hours of full power a day. We have sunlight for 10-12 hours most of the year, but you only get peak power for 4 hours, 1/2 power for another 4, 1/4 power for the other 4.
When you actually live on solar power, you learn how it really does in a real world situation.
Also, you add a tracking array, and you can gain an extra 25% of power per day. There are other methods for gaining efficiency, like MPPT controllers, etc, but I won't go into that too much here.
You said $0.5/W, so we'll stick with that; the cost of the panel is $166/panel in bulk, for a cost of $0.72/W;
yeah, because you are paying Canadian markup, shipping to Canada, etc. Buying in bulk in China is considerably cheaper. And when you are talking a data center, you are talking serious bulk. That price is for 20 panels, which would be great for a house or 2, but not for a data center. Now, imagine buyng 2,000 panels straight from the factory in China.
For a system like this, you want grid-tie, to avoid needing batteries, so at night, you pull from the grid, and in the day, you feed to the grid. It is a standard installation, and even in China, you can get it done quickly and professional at little cost.
If you are seriously interested in this stuff, PM me, but I won't clog this thread anymore with solar talk. I'm just saying it is certainly possible, and his ROI is lower than the figures you are using.