This can be solved with for example IPFS. When you download a file you also make it accessible to others.

This means that reverse scaling effect kicks in. The more popular the file is the less people access it from original source.

Which is why you want content-addressable decentralized storage, so the content gets distributed either peer to peer or through public caching infrastructure that's operated by ISPs, just as IP router are nowadays.

I'd say the main thing is people's inability to reason without serialisability.

Without the serialisation of actions that centralisation gives, it's much harder and hence more expensive to create things.