Neither. What I meant is that clouds need to be Tinkertoys that can be used to assemble services with no single point of failure. S3's Achilles heel is that it's all Amazon. If Amazon has a widespread failure or decide to stop doing business with you, your service is instantly, irrevocably dead in the water.
If you're able to spread your storage among multiple providers as S3 does among Amazon's own machinery, you're no longer at the mercy of one provider. Contract with P1, P2 and P3 to do your raw block storage. Contract with P4 and P5 to run the storage service against what you have stored at P1, P2 and P3. If P2 goes poof, you keep running, contract with P6 for new bulk storage, tell your storage services at P4 and P5 you have new space at P6 and it gets loaded up, same as Amazon would do if a machine holding your data fails. If P4 goes away, you fail over to P5 and contract with P7 to run your aggregation service.
I wrote a whole article about this topic a couple of years ago, including some discussion about what will have to happen for something like this to come to fruition. Not wanting to be a blog link ho, I didn't post it here.