How would the inefficiency of the modern web be alleviated by replicating the same data on more nodes? That _increases_ wastefulness, not decreases it.
Replication is cheap. When the data travels over the wire, it's being constantly "replicated" at every hop (data doesn't really "travel", it's just being copied over and deleted afterwards). When you send data halfway around the planet and back, you're doing a lot of copying. IPFS model makes it so that after you do that, other people around you don't have to do the same long roundtrip to get the same file.