Is dApps not just expensive peer-to-peer network?

First of all, there are different categories of DAPPs leveling on different blockchains and different layer solutions.

Peer to peer based DAPPs that allow interchange of funds,tokens or other information are one of those types.

Another thing we must understand is that μTorrent and platforms like PirateBay had already been using decentralized servers before the advent of present day blokchains as they did this in late 90s.

So what new has been added now that only Decentralized Apps can solve?

Firstly, the problem of (PoW). Any asset fundamentally is based upon trust. The monetarily value of anything comes from trust and security. Fiat currency is secured by advanced Bonds, Securities and other militial contracts.
Similarly, to provide value to the decentralized units of currency, Proof Of Work is an effective mechanism that basically is a long competition between smart contracts to solve a cryptographic puzzle first. On the basis of this “mining” is done.

This requires a combination of  packages and architecture at different layers with millions of nodes interacting altogether.
This alone cannot be solved by peer to peer networks.

Secondly, Staking Pools and Yield Farms and Liquidity Pools are other DAPPs that require more architectural packages and APIs other than peer-peer network. There must be Layers and webpacks with a lot of combined dynamic stacks over a Blockchain to run it in efficient amount of time.

The advent of Ethereum 2.0 recently has significantly decreased the runtime and processing by 100th fraction  while staking via PoS because of L-1 and L-2 additions over it which has been done by further additions over existing reinforcements.

A peer to peer network is not feasible for larger scalability of stacks and layers of solutions over any decentralized networks

The entire idea of solutions like Ethereum, Arbitrum and Fantom is to reduce dependency over peer to peer networks and build more solutions and dApps over the Blockchain themselves.

With Sharding Tools, Arbitrage By Time Stakers and Insurance Mining, the peer-peer network itself is not sufficient enough as these decentralized solutions require more packages and segmenting APIs to handle them. This has created the advent of entirely new marketplace of Web 3.0 solely focusing on further developments of such protocols and packages that can reduce the cost and time gap burnout .

In short, basic Decentralized Applications that solely focus on information exchange can be based over peer-peer network but the present day need of Web 3.0 arch requires more protocols and solutions to increase efficiency and decrease burnout cost/fees.