what aws is not
In 2004, SQS and AWIS beta-ed.
In 2005, MT beta-ed.
In 2006, S3 and EC2 beta-ed.
From there, the pace of releases has skyrocketed (something we should put value on). AWS started by turning basic computation services into utilities. They've since done the same to a wide range of technology capabilities--dozens of services, hundreds of options, a combinatorial explosion of capabilities. So far so that we could reproduce all the functions and services provided by any data center anywhere.
That's where AWS is. AWS is not a commodity, though specific AWS services may become commodities. AWS is not basic computation services. AWS is not just for startups or web2.0 or mobile or small shops or transient projects or marketing or unregulated.. etc.
AWS is the successful utility-isation of ever more, and ever more valuable, technology services.
They are building the AWS of next year or further out through utility-ifying whatever it is that their ecosystem (customers included) is telling them (through behavior) is worth paying for.
To really compete, you'd have to: match the ecosystem play and exert margin pressure. The former you could do through co-option--which would require taking over the service supply chain--or through drawing your own ecosystem to some core differentiation (e.g. live migration on GCE or seamless public/private experience on Azure). The latter can only be afforded by a few organizations (Google and Microsoft).
Hat tip: Most of the thought above is a direct result of, or informed by, Simon Wardley.