Quote from endsongs:
Sorry but your land price estimate is way too high. The link below shows there is a lot of private land availabe for ~$200 to $1000 per acre.
http://www.landwatch.com/New_Mexico_land_for_sale/500_Acres?
So, there are ~640 acres per sq mi. Assuming a land cost of ~$500 per acre this gives a total land cost of about $964 billion. Also, I'm sure a lot of the land used would be borrowed from the gov't, which would also reduce the cost.
Yes, my price was the cost of the entire project of building enough solar field to supply the US with electric power. Not just the land cost. Funny, I actually used $500/acre in my estimated cost of the project and that was just off the top of my head.

The bulk of the cost is in the solar panels themselves. Even after dramatic price reductions over the past two years they are still quite expensive relative to coal/oil/NG. The panels alone would cost about 4.5 trillion, and then you have to figure in racks, inverters, wire, roads, trackers, etc. $10 trillion would barely get us there, and that is only the US. Then you have to consider that while we are making huge progress with solar panel efficiency and cost, there is still almost no ability to store large amounts of electricity, which I believe is the point that Hydroblunt was trying to get at (although he could've been a bit more civil about it).
To switch over to solar would take a complete revamp of how we perceive our energy usage, mainly because of the on-demand nature of the way we use electricity. As a yearly average, solar panels only hit the equivalent of full output about 5.5 hours per day, year round. More in the summer and less in the winter. Cloudy locations are horrible and sunny locations exceed that estimate, but that is the average. Current economies of electrical transmission make it feasible to transmit a max of about 3,000 miles before the law of diminishing returns starts to hurt too bad. Therefore, at any given location you could only expect to draw power from panels located within a couple thousand miles. This does effectively lengthen the daily production and (contrary to popular opinion) on average most electricity is used during the day.
So imagine a system that worked like this. Solar panels were littered about everywhere, on rooftops and such. These were mainly grid-tie systems w/ backup on-site storage for emergency use only. So for example a home would be fitted with enough solar production to supply itself, but it didn't actually use that power, but instead fed it into the grid under a net-metering contract with the power company. This is usually what happens currently anyway. But in this future scenario, a total of all panels within a 2,000 mile radius supplied the vast majority of all daytime consumption. Peak morning and afternoon demands are somewhat compensated by the ability to draw power from locations that are in a peak production time zone, but haven't yet reached peak demand. Current power plants would then serve the purpose of simply satisfying abnormal demand during severe weather events and a short nighttime period. Under such a scenario they might only see about 25% of current production rates.
We would then have decentralized the power production such that it would be much more difficult to cripple the system, and a large portion of the buildings would be able to operate on backup power storage during emergencies. I'm sure as technology progresses one of two things will happen. Either storage capability will increase, or the ability to cheaply transmit power over much longer distances will become a reality. In the mean time, this proposed situation is still much better than our current one. I think it will become a reality once solar cost drops to about 30% of current prices.