Walrus has recently become a project I keep studying. To be honest, I’m not attracted by its story, but by the fact that it hits a pain point that no one pays much attention to.

Currently, the consensus, assets, and execution in Web3 are quite mature, but there’s a reality that no one likes to talk about: the real resource consumption isn’t from the transactions themselves, but from data. Images, videos, model parameters, game assets, social content, AI intermediate results—once these are on-chain, costs explode; if not on-chain, you can’t enjoy Web3’s verifiability and persistence.

Walrus aims to address this long-overlooked gap. But unlike those decentralized storage projects that say "I can store," it truly asks "how to store, for whom to store, and in what scenarios." From the very beginning, its design didn’t treat individual objects as a few KBs, but instead focused on scales of 20-50MB or even larger. This is simply not feasible in most protocols—they’re still stuck at storing JSON and small files.

The logic behind this is actually quite clear: Web3’s true explosion won’t come from transaction volume growth, but from a surge in content volume. Future competitiveness depends on who can efficiently handle large-scale data.