Storage has traditionally been dominated by veteran projects like Filecoin and Arweave, but recently an interesting alternative approach has emerged. There is a project that doesn't follow the traditional IPFS route but instead builds a storage layer on Sui chain, specifically tailored for AI data and large files.

This project is Walrus Protocol. Its current token price is around $0.15, with a circulating market cap of $238 million. While its size may not seem particularly eye-catching, its design philosophy is quite innovative. Most storage protocols treat all data equally, but Walrus's idea is to optimize specifically for binary large files and rich media content. It's worth noting that AI model training now involves datasets of hundreds of TBs, so this targeted optimization approach is quite pragmatic.

The development team behind it is Mysten Labs— the core team behind Sui chain. This identity naturally provides opportunities for deep integration between Walrus and Sui's high-performance architecture. Currently, the total storage capacity has reached 4,167TB, with an utilization rate of about 26%. This ratio still has considerable room for growth. Compared to older projects whose storage capacities are already saturated, Walrus is clearly still in its growth phase.

From a technical architecture perspective, Walrus uses erasure coding—while this technology isn't exactly new, the key point is that it enables programmable access control on Sui. This is worth noting. Traditional storage access permissions are often static, but programmable control means storage policies can be flexibly adjusted, which will be practically beneficial for complex AI applications.