This is a fantastic question for pretty much every scaling solution out there — as the initial engineering work on rollups finish, many of the fundamental scaling problems re-emerge on L2. Right now, our system's hardware requirements are very similar to L1 mainnet, but the state is growing.
There are two solutions in the future: statelessness, and block-producer/verifier asymmetry. Statelessness (and related concepts like state expiry) has been under active research in Ethereum for years, and we've recently started our own contributions with a new stateless Ethereum client [1]
The other part of the solution is to leverage asymmetries between the hardware requirements of block producers and verifiers. TLDR: this lets you have high HW requirements for sequencers, but still secure the network with laptops. Vitalik recently wrote about this; you can read that here [2]
There are two solutions in the future: statelessness, and block-producer/verifier asymmetry. Statelessness (and related concepts like state expiry) has been under active research in Ethereum for years, and we've recently started our own contributions with a new stateless Ethereum client [1]
The other part of the solution is to leverage asymmetries between the hardware requirements of block producers and verifiers. TLDR: this lets you have high HW requirements for sequencers, but still secure the network with laptops. Vitalik recently wrote about this; you can read that here [2]
[1] https://twitter.com/ben_chain/status/1488275978983915523?s=2... [2] https://vitalik.ca/general/2021/12/06/endgame.html