I find Amazon more interesting because they are exposing the plumbing required to do sites in mass.
Google's AppEngine is much closer to the world of "Digital Sharecropping" that Jessie has been describing. Amazon has some of this in their API's, but I would bet that the more successful pieces to Amazon Web Services are the ones furthest from the share cropping model.
S3 and EC2 have relatively little tie in to them. You can end up with a physical addiction to the services but the mental addiction to a framework does not exist. S3 is just storage and EC2 for most is just a hosted Linux image.
I would also suspect that AppEngine solves two problems for Google. Those being that Google will eventually need a way to do department applications for companies growing up using Google Applications, and the second being that App Engine is Google's attempt to outsource their 20% projects to the world.
If Google can get the next Twitter to build itself on AppEngine then they will have a success and have a better chance of lowering their rejection rate for acquisitions. Looking at the failures around Dodgeball and similar systems it stands out that Google could not absorb applications into their "one way is the right way" methodology. YouTube is a success but then they are essentially what they were from the beginning (for instance they continue to use MySQL over Big Table).
Amazon though? They may someday acquire a company that is running in their cloud, but it does not look like they are concentrating on this. Instead they have a very open system, in terms of getting data in and out, that they are publishing to the world.
Think about the success of open source. You learn more and you see a little further every time you gain a new user. Amazon has the radical approach to let the world test their systems (and pay for it while doing so). They build out and subsidize their main businesses infrastructure through AWS.
For these reasons, when I look at these two companies, I find Amazon's approach more interesting.