By 2026, a consensus is forming in the Web3 world: data is the strategic resource of the new era, and decentralized storage is the key to this resource.

Speaking of the storage track, Filecoin competes for scale with its vast node network, while Arweave attracts investors with its "pay once, store forever" story. But how does it actually perform in practice? Costs are too high, data recovery is slow, and there's little flexibility for programming—none of these pain points have been solved.

In this sea of fierce competition, a new player is beginning to emerge. Walrus is a project nurtured by Mysten Labs, the parent company of Sui. It activated its mainnet in March last year, raising $140 million (led by Standard Crypto, with participation from a16z crypto, Franklin Templeton, and others), and at one point, the market valued it at $2 billion. When it comes to funding and valuation, these figures are already quite impressive in the storage sector. But Walrus didn't rely on stacking parameters or hyping token concepts to generate buzz; instead, it took a completely different approach.

Its core strategy is actually very restrained: it doesn't try to be a jack-of-all-trades, all-encompassing platform. Instead, it focuses on two high-value, high-willingness-to-pay scenarios—AI model training and inference data storage, and real-world asset RWA on-chain.

Why are these two areas worth betting on? With the rise of AI, data volumes are exploding exponentially; a training dataset can easily reach terabytes. Small and medium AI teams initially wanted to use Filecoin, but after doing the math, they were scared off: redundancy factors start at 25x, and storing 100GB of data for a year would be prohibitively expensive. This is the opportunity Walrus saw. Compared to the "big and comprehensive" fantasy of decentralized cloud storage, Walrus chooses to excel in vertical scenarios, becoming the most critical infrastructure piece in the Sui ecosystem.