In June of this year, Walrus launched the Quilt upgrade. The official team was confident at the time, claiming that this feature could help users save over 3 million WAL in storage costs. It sounded pretty impressive, but if you actually look at the on-chain data, you'll find that things are not that simple.

Walrus's revenue performance directly collapsed. In Q2 2025, it still had $380,000 in protocol revenue, but by Q3 it dropped to $30,000, and in Q4 it continued to decline to $18,000. By Q1 2026, it was down to just $271. Dropping from $380,000 to $271 is an explosive decline. Even more strangely, this timing coincides exactly with the launch of Quilt.

Why did Walrus develop Quilt? The technical logic is actually quite clear. Before Quilt, Walrus encoded each file individually into a blob—even small files of a few KB had to go through the full erasure coding process, ultimately generating dozens or even hundreds of KB of redundant data. This was a disaster for scenarios involving massive small files like NFT metadata, social media posts, and configuration files—storage efficiency was abysmally low, and costs were extremely high.

Quilt's solution was to bundle multiple small files into a large blob and then encode them in batches. This significantly reduced redundant data and greatly lowered storage costs. From a purely technical perspective, this optimization makes a lot of sense. Filecoin and other storage projects are also using file aggregation techniques.

But here’s the critical point: Walrus’s charging model is based on the amount of encoded data. The more encoded data users store, the more revenue the protocol earns. Quilt directly reduces the volume of encoded data, which naturally cuts into revenue. While technically optimized, this approach effectively undermines the business model.