The next step in cloud computing?
The economics of cloud computing is based on the assumption that cloud providers can get better deals on things like servers and power than smaller businesses can because the cloud providers buy things in huge quantities. The best discussion that I've seen of this is Microsoft's "The Economics of the Cloud" (PDF). This white paper claims that the biggest area in which cloud providers have an advantage is in the area of cheap power. It also says that this is becoming the biggest component of the cost of IT.
So an obvious question to ask is this: at what point should we expect to see cloud providers generate their own power?
A typical power plant might produce roughly 1,000 MW, while a typical server might consume something like 500 W. This means that a single power plant can probably keep about 2 million servers running. (Maybe it's really more like half that number once you look at how much you need for air conditioning, etc.)
But industry estimates already say that Google isn't far from having that many servers. They're probably over half way there. (Other cloud providers seem to be behind by a factor of 10 to 20 or more.) So it shouldn't be too surprising if we start to see cloud companies thinking about generating their own power. Google might already be doing this.