The real drive behind this might be the need for power in the military (though that does not make a lot of sense as the power will be off during the night, and the batteries the military would need to store power for use overnight would likely be bigger than the oil fueled generators they replace). It might be the military sees this as a way to get a new weapon in the sky, consider their recent work with the "pain ray", a microwave gun that causes intense pain by stimulating the nerves of the skin, perhaps such satellites (which use microwaves to relay their collected power to the ground) could do double duty and be used to cover large areas of a battlefield with pain rays, thus, knocking enemy troops out of action prior to an attack. 
Currently I think a virtual private server (VPS) solution is the best bet for those who need to start small, and while EC2 has some advantages its pricing is currently a lot higher. Going the VPS route has some scalability, some vendors (such as linode.com) offer about a 10:1 scaling ratio in features across their offered services.
Once you have maxed out a typical VPS vendor's offerings you are in the price range of a single dedicated server so the migration path could be continued by switching to a dedicated server or by getting your own hardware and perhaps co-locating it. Doing this could add about a factor of 5 to the scaling curve, so in total, the virtual and dedicated private server approaches should allow you to scale your application about 50 times without having to rework the architecture or selected technology. Once you have grown to encounter those limits you are probably leaving the domain of the startup, so its probably time for a rethink anyway.