TL;DR
- Walrus will launch Quilt to automate the packaging of small files and reduce storage costs on the SUI network.
- Quilt will bundle up to 660 objects per batch and cut storage costs for 10 KB blobs by up to 420x.
- The Quilt API will integrate metadata and direct access to each file, making it ideal for AI, event logs, and NFT collections.
Walrus will expand its storage capabilities with the launch of Quilt, a solution designed to optimize the management of large volumes of small files.
Since the launch of its mainnet in March 2025, Walrus has stored over 800 TB of data across 14 million blobs. However, storing thousands of small files required developers to manually bundle them to reduce costs. With Quilt, that task will be eliminated, as the system will automatically group files into efficient blocks.
How Quilt Works
Quilt will act as a batch storage system. Instead of storing each file individually, it will bundle up to 660 objects into a single unit, reducing encoding overhead and gas fees on SUI.
Storing small files with Quilt will lower costs by up to 106x for 100 KB blobs and 420x for 10 KB blobs. Despite being bundled, each file will remain individually and quickly accessible, with no need to decompress the entire batch.
The Quilt API will allow this functionality to be integrated into any application using Walrus. Developers will be able to add protocol-native tags or metadata to efficiently organize and retrieve content. This will be especially useful for systems generating AI data, event logs, or non-fungible token collections. Quilt will also offer a responsive experience by enabling low-latency access to specific files.
Walrus Proves Its Tech with Real-World Projects
During its testing phase, Quilt will already prove its value in real-world use cases. Tusky, a decentralized storage platform, will reduce the overhead of managing small files during high-demand periods. With Quilt, the application will maintain consistent performance while minimizing gas expenses. Meanwhile, Gata will use Quilt to compile AI training datasets, allowing each agent to access data fragments without manual file grouping.
With Quilt, Walrus will complete its offering for handling files of any size. From now on, creators will have a single platform to manage everything from multi-gigabyte media to files just a few kilobytes in size. Quilt will let developers focus on innovation while ensuring efficiency, individual control, and reduced costs
