The AI-specific data storage platform Shelby, jointly developed by Aptos Labs and Jump Crypto, is now open for public preview access. Built on DoubleZero's dedicated fiber optic network, the platform aims to address the common challenges of read efficiency and data provenance in AI workloads. Developers can integrate Shelby in advance through the testnet to validate its performance in real-world scenarios, with the official version expected to launch in 2026.
Traditional object storage systems are designed to handle high-frequency write scenarios, while AI applications are quite the opposite—single user interactions can trigger dozens of parallel reads across regions and datasets. To tackle this challenge, the industry typically employs solutions like data replication, caching layers, and multi-copy synchronization, which not only increase storage costs but also complicate data consistency management. More critically, when copyright-protected or user-authorized data is included in model contexts, there is a lack of reliable evidence chains to prove the data's access time, source, and usage rights. Most teams rely solely on temporary logging, making it difficult to meet compliance audit requirements.
Shelby addresses these two major pain points from the ground up. Its technical architecture is supported by three core pillars: DoubleZero provides a dedicated fiber optic network spanning over 30 cities across five continents, ensuring sub-second stable transmission; Jump Crypto contributes its high-concurrency, low-latency storage system optimized for quantitative trading environments, catering to AI's burst read demands; Aptos serves as a decentralized coordination layer, generating immutable cryptographic certificates for each data read through blockchain technology, anchoring data ownership and usage rights, and achieving end-to-end verifiability.
The first batch of institutions participating in the preview includes NBCUniversal, Metaplex, Story Protocol, and Pipe Network. Notably, NBCUniversal, as a leading global media group with a vast library of copyright-protected audiovisual content, marks the extension of this technology from the crypto sphere to the mainstream content industry, validating the immense potential of verifiable data infrastructure in digital rights management and compliance operations.