Via groundcover.com
The decentralized storage protocol, built with Jump Crypto, promises sub-second reads and 70% lower costs than traditional cloud providers
Every AI company eventually hits the same wall. The models get smarter, the compute gets cheaper, but the data layer, the plumbing that actually moves training sets and inference results around, remains stubbornly expensive and slow. Aptos Labs thinks it has a fix.
The Layer 1 blockchain developer unveiled Shelby on June 24, a high-performance decentralized object storage protocol built in partnership with Jump Crypto. The pitch: sub-second read speeds, built-in monetization for data providers, and egress costs roughly 70% lower than what traditional cloud services charge.
What Shelby actually does #
Object storage is the backbone of how modern applications handle unstructured data, everything from images and videos to massive AI training datasets. Today, most of that storage runs through a handful of centralized providers. Amazon S3, Google Cloud Storage, Azure Blob. They work well, but they come with vendor lock-in, opaque pricing, and egress fees that can quietly eat into margins.
The sub-second read capability is the headline feature. Most decentralized storage solutions struggle with latency, which makes them impractical for real-time AI workloads. Shelby’s architecture leans on Aptos’s low-latency transaction finality to keep retrieval times competitive with centralized alternatives.
Cloud egress fees, the charges you pay every time data leaves a provider’s network, are one of the most hated line items in any engineering team’s budget. Shelby claims a 70% reduction in those costs.
The business model and token economics #
One notable design choice: Shelby doesn’t have its own token. The protocol uses APT, the native token of the Aptos network, for all gas fees and transactions.
Shelby also bakes in read incentives at the protocol level, meaning data providers can monetize access to their stored objects directly. This opens the door to AI data marketplaces where companies can sell or license datasets without intermediaries.
Use cases extend beyond pure AI workloads. The protocol supports pay-per-view content distribution, decentralized application backends, and integration with existing Web3 infrastructure. Projects like Metaplex and DoubleZero have been noted as potential integration partners.
Jump Crypto’s involvement and what it signals #
The partnership with Jump Crypto adds significant credibility. Jump’s crypto division, an arm of the Chicago-based trading giant Jump Trading, has previously contributed to core infrastructure across multiple blockchains, including building Firedancer, an independent validator client for Solana.
The Aptos blockchain itself provides a natural foundation. Built by former Meta engineers who worked on the Diem project, Aptos uses the Move programming language and a parallel execution engine designed for high throughput.
Where this fits in the competitive landscape #
Decentralized storage isn’t new. Filecoin, Arweave, and IPFS have been working on this problem for years. But most of those solutions were designed primarily for archival storage. Shelby is positioning itself differently, going after the hot storage market, data that needs to be accessed frequently and fast.
Shelby has already moved beyond its early access phase and into private production, with reports of customer onboarding emerging.
For APT holders, the implications are straightforward. Every transaction on Shelby generates demand for APT tokens. If the protocol gains meaningful adoption among AI companies, it creates a new source of organic demand for the token driven by actual usage rather than speculation. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our