Amazon’s EC2 and S3 compete in a growing and changing marketplace; not really a marketplace of competitors, more of competing methodologies.  

• Google would have you write Python (and now maybe Java) and compile it to their cloud, charging you when you really put the pedal to the HoTMetaL.  
• Microsoft’s Azure makes a cloud out of the pile of machines you already control, but doesn’t really change the landscape around paying for these things granularly.
• GoGrid, Rackspace’s Mosso, MediaTemple’s Grid and a host of others provide semi-painless scale to simple web hosting, but miss the boat around applications other than surviving a slashdotting and making AdSense cash.
• Xen, Hadoop/Nutch/Mahout, and a host of other applications provide you frameworks, but require real deal, no BS serious IT support to get to a place where you’re extracting value.

Amazon’s model is someplace between all of these.  You need a little more tech-savvy (or you can buy it in the form of an AMI), but you retain the flexibility of language, OS, and incremental payment control.  You get whole machine “instances” so you can develop the way you would normally (somewhat), but you might pay slightly more for it. Some of the new elements of their model (hadoop images, reserved machines for deeper volume discounts, etc) position them well in the enterprise space. Storage via S3 provides a valuable abstraction layer as well. Ideas that might furhter distance them from the pack include stronger free trials of limited performance AMI’s, case studies on known IT challenges (getting SQL server to scale out rather than up, reducing cost on temporary test environments, etC) as well as strong specs on performance around disk I/O (how fast really is S3?) and network speed between nodes (can you pay for neighbors in racks?) as well as inbound/outbound network performance from the EC2 hosting environments. Amazing where we are today!  Welcome to the Cloud!