A couple of days ago, I was chatting with a developer working on Web3 games. He mentioned that his biggest headache is storage issues. Using traditional cloud providers? The costs are terrifying, and you have to hand over all your data for others to manage; want to use decentralized storage solutions? Stability is risky, and the prices aren’t exactly cheap either. It wasn’t until he mentioned that he’s been paying attention to Walrus recently that his whole demeanor changed: "I heard their prices can even beat AWS?"

This observation made me start to think. Why does Walrus seem to have a chance to reshape the entire storage sector? Ultimately, there’s one reason: it’s truly cheap. This kind of affordability isn’t achieved by burning investors’ money, but stems from the advantages of the underlying technology architecture itself.

So here’s the question—why have existing decentralized storage solutions failed to reduce costs?

Take Filecoin as an example. This project is indeed a representative of decentralized storage, but it has a fundamental design limitation. To ensure data security, the entire network uses a full replica redundancy mechanism—meaning, when you store a file, the system makes multiple complete copies across various nodes in the network. Security is guaranteed, but the cost is enormous hardware resource consumption. It’s like sending a letter but hiring ten postal workers to walk ten parallel routes; of course, the costs can’t be lowered. In the end, it even becomes more expensive than some centralized cloud services, which is quite ironic.

Walrus, on the other hand, adopts a completely different approach. They introduced erasure coding technology, which is quite clever. Simply put, it involves breaking a file into several fragments and establishing relationships between them through mathematical equations—so you only need a subset of the fragments to fully recover the data, without storing complete copies. Security is maintained through redundancy, but the costs are significantly reduced. This is what’s called a "dimensionality reduction attack."