Ask anyone involved in standing up a public cloud service and you'll hear the same story. The technology is relatively the easy part. The hard part is defining the business model aspects of your cloud service offering, more specifically defining sales models and pricing that works.
Often you'll hear a variety of buzz words thrown around like econmies of scale or low margin / high volume business when talking about Cloud Computing, but for the most part competing based on capital expense and volume isn't a reasonable approach. (We'll just out spend our competitors, ie. Microsoft spending billions on new data centers) Most revenue models break down into two basic groups, technical value (features) and cost (price / margin). As I've written before, value is about solving problems, but competitive pricing is also important.
Starting at bottom you have the freemium model, arguably one of the most useful approaches. This a model where you give away your service in hopes of monetizing at some later point or possibly selling ad space. I'd classify this as a high volume approach. Lots of users, some may pay -- maybe. The Free approach to service generates lots of users as well as lots of up front operational costs. So you'll need to consider your cost per sale ratio (CPS) closely. How much does it cost me to convert a free user into a paying customer? This converstion concept was made popular in the SEO realm where sites would calculate their ad prices based on the number of sales transactions an ad could potentially generate. Basically it was ROI based pricing. It's also the main driver behind the commerical open source movement. But today we're not talking about selling traditional software, but instead selling cloud services (SaaS, IaaS, PaaS, etc,).
Another freemium approach is the free introductory service or a loss leader where a service is sold at a low price (at cost or below cost or free) to stimulate other profitable sales. If you've ever been to a retail store, you'll recognize this sales promotion tactic. Free ballons for the kids, cheap beer for Dad etc. What's interesting from my vantage point is a number of Enomaly ECP service providers have used this approach with great success. One of the more successful created a free "Beta" period which generated hundreds of user signups. This both helped test the cloud IaaS service & platform as well as provided a large pool of interested, and more importantly targeted customers to convert. And convert they did -- with more than 70% of the free customers ending up paying for the service after the beta phase was complete. I jokingly refer to this approach as the crack dealer sales approach. Get them hooked early and quickly.
Other approaches could include deferred payments, (we won't charge you unless you use it, or won't charge you for the first 90 days). You need to realize that users like free. But customers are willing to pay for something they like or see value in.
Which brings us to how much should you charge for your cloud service? Assuming you have competitors, (if you don't than you probably don't have any market for your service either) you'll want to likely be in the same ball part for the same basic service as your competitors. This will probably be an entry level service with little differentiation. The important aspect to pricing will be in the areas where you're different. Maybe you offer a premium storage or load balancer, or advanced security or something I haven't even thought of. The more value you provide the more you can charge for your services. Also the key metric is ROI and making sure you can easily demonstrate it, hopefully without a sales person - low or no touch sales are always better. These are areas that you can charge significantly more for because, you're costs are 'obviously' higher, and hopefully it's a better value than your customer attempting to create the service themselves or to buy from someone else.