I think the idea of Rate Limiting usage as a business model is very interesting. I recently read that Google is going to begin charging for using their Prediction API for a fee monthly plus usage. Another popular example is Twilio, which charges based on usage of their API. I could be using this term incorrectly, but this is how I see them controlling pricing and their costs. SendGrid is yet another example that controls pricing simply based on how much you use the service. There are many other examples.
In these examples, Companies are essentially lowering the barriers to entry to zero, even less than free in the case of Twilio who gives you a $30.00 credit, to get you started. This make using an API appealing to developers who may spend the time building an application. Once up and running, the idea is that if the app is getting usage and traction it should pay for access. If its getting serious traction then it could be making money, hence charging for the access to the underlying services.
By having developers build on their services, these Companies and others make switching costs a little bit higher (mainly time and ramping up to a new service) but a compelling way to attract people to use their tools. It is in some cases a freemium model at its best. Without having an “unlimited” plan at the top tier, these Companies can scale up their pricing to as large as the enterprise using their product. I am sure most have special plans available, but the initial possibility of infinite cost is there.
With CPU cycle costs dropping, and overall hosting costs getting lower – value add algorithmic products or just plain products hosted in the cloud can serve a big need. They solve problems faster than rolling your own solution, and have initial costs so low it doesn’t make sense to go elsewhere.