Analogies are useful things. It helps us predict the future for something that's new and it helps us form opinions when there is still a great deal of uncertainty. The real power in any particular analogy is how well it helps you draw inferences. Cloud computing is still groping around for analogies and a few have popped up. The 'cloud' moniker itself is an analogy. It mostly works for the end user who has no idea how this stuff is built. But for anyone who has ever walked through a major datacenter, it couldn't be further from representative. Cloud computing involves a lot of heavy equipment and big machines. It's anything but vaporous.

Some look to power and other utilities. This angle has some benefits. It helps you think about locality and the network. You get benefits of scale. It involves wires. But that's about as far as it gets. The problem is that electricity is a commodity. Compute power is not a commodity. Now I understand the people who use this analogy feel that computing somehow should be commodity. Of course everyone who consumes a resource has an interest in making that resource a commodity, But that doesn't make it true. There is more innovation in cloud computing than simple operational efficiency and cost cutting. The combinations of platforms, development environments, and application ecosystems are anything but commodity.

Others like to think of macro-economic forces.  James Hamilton has a post about cloud computing that relies heavily on economies of scale to make his point. I like this way of thinking. But his conclusions are less than satisfying. The problem isn't with using economics, it's with using naïve economics. We are learning more and more that people aren't rational in the economic sense either individually or in groups. This isn't something to complain about or try to change. It's simply a fact that needs to be understood. People optimize for more than just money. There are government regulations and the very human desire to specialize or have greater control. Besides, economies of scale don't extend indefinitely. As the platforms mature, many organizations will be large enough to get the benefit of scale without sharing resources.

I like to think of cloud computing like healthcare. There are great benefits of scale for insurance providers. A big plan can amortize risk over a great number of users and negotiate better prices. But that doesn't mean that all insurance companies will consolidate. Some employer provided insurance plans don't even share the subscriber pool with non-employees. For example, Microsoft offers fantastic health benefits and it's entirely self-insured. It's big enough that it gets the benefit of scale from its own subscriber base. Only the administration goes through a third party. Similarly there are plenty of companies that are big enough to get the scale benefits of cloud computing but still want to run things themselves. There are also other factors as well. For better or worse, governments get into the action with healthcare. Some governments control everything, others fund and regulate private markets, others still are a hybrid. Just because you may be convinced that one model is best doesn't mean that other models won't emerge as well. Just as certainly some governments will be involved in the evolution of cloud computing. Some will run their own. Others will be content to regulate. Others still will be largely hands-off. Government involvement in cloud computing will have both consolidating and fragmenting effects. There are plenty of privacy and other laws that easily trump the economies of scale arguments.

Economies of scale are powerful driving forces in cloud computing, but they aren't the only ones. I expect to see many business models emerge. There will be hosters, enterprises, and governments running cloud platforms. Some providers will compete on cost, others will offer special services like high IO storage systems, low latency interconnects, or  dedicated hardware. Some will integrate vertically and provide soup to nuts solutions for particular industries. Others will specialize horizontally and work with a wide range of partners. Some enterprises will choose to run the whole thing themselves using the same underlying cloud infrastructure. There will be new business relationships formed and new supply chains created.

Some of these business models will succeed. Others will fail. If you like intellectually pure arguments, you may still be convinced that private clouds are a transitional state and that in the long run there will only be a small number of public clouds. That may be, but as Keynes famously said, "In the long run, we are all dead."