Many in solar power focus on efficiency. For every watt of sunlight, how much of it can we turn into electricity? 10%? 20%? Billions have gone into researching more efficient solar cells, with substantial progress over the last thirty years.
But efficiency is the wrong thing to focus on. Who cares about watts per square meter? Square meters are free. We have a bazillion of them – we can use as many as we want. The key metric is watts per dollar. Electricity buyers care almost exclusively about cost. Watts per dollar determines whether they use solar or coal. If a new panel is 10% more efficient, but double the price, it’s useless.
For a cheap solution, how about this: Take a bunch of desert. Cover it in a layer of insulator, then (on top) metal, black paint, water, and glass. When the Sun comes up, it will heat the paint very quickly (~20 – 100 C / hr.), causing the water to become high-pressure steam. With double-paned glass and good insulation, the heat will be trapped well enough for steam to reach many hundred degrees C. That steam can then run an old-fashioned steam turbine.
At 250 C, maximum (Carnot) efficiency would be about 40%. In practice, steam turbines aren’t very good, so you might only get 10%. But even at 10%, time-averaged power output would be ~25 W/m^2, or $100/m^2 at current $4/W prices. And there’s no way bulk insulation and low-quality glass panels cost $100/m^2.
The Rationalist Conspiracy 
http://en.wikipedia.org/wiki/Solar_updraft_tower People are already working on very similar ideas. I like it.
Like many other great ideas, it’s been thought of before. I don’t know why it’s not being used much.
I did a bit more clicking. Yes, you can make solar thermal plants out of cheap materials, but that’s not the issue. It’s still expensive to hire people to build the things and maintain the engine that your solar heat source is driving. Apparently, photovoltaic cells are actually beating solar thermal on overall price per watt.
As a follow-up to Douglas’s comment, perhaps the real bottleneck to cheap thermal solar power is we don’t have sufficiently cheap and robust robots. This, in turn, is a software problem.
“When all you have is a hammer, everything looks like a nail.” – Abraham Maslow
Given the very large fraction of American talent focused on software (explored by me earlier), people try to define almost everything as a software problem. Of course, not everything is, and in this case it’s especially silly.
Suppose, for a moment, there was perfect code for running any manufacturing robot, just floating around for free on the Internet. Now, who will build all of these robots? Who will design all the blueprints? Who will finance the factories? Who will deal with the safety and regulatory hurdles? Who will optimize the design to be cheap enough to replace human workers? Who will run the reliability testing? Who will handle the upstream suppliers? All of that stuff doesn’t just happen by magic, and will probably be far harder than just writing some code.
That won’t work as described. Water at several hundred degrees is at a pressure of tens or even a hundred atmospheres. You cannot hold that back with a flat barrier, especially not low quality glass paneling. Maybe liquid salt or something, but then it’s not so simple.
To contain your hot liquid, you must use round pipes, in which case you need solar concentrators (mirrors), which then require re-aiming every few days to follow the sun. This is in fact done, but not so easy as to dominate other forms.
But the temperature during the night can fall very low afaik. Meaning large amount of water (thick layer) would never heat up properly during the day, making the entire project inefficient.
How about switching to PoS or PoS hybrid instead?