Aptos moves Shelby into private beta with AI workloads

Shelby, the decentralized data infrastructure project built by Aptos Labs, has entered private beta with its first customer workloads now running in production, the company announced. 

Three companies across enterprise AI, distributed 3D rendering, and physical-world spatial data are the first to go live on the platform.

Aptos is a Layer 1 blockchain originally incubated at Meta. Shelby was first announced as a high-performance data layer for Web3, essentially infrastructure that sits beneath applications and handles how data is stored, moved, and made available. 

Since then, the project has sharpened its focus toward AI and other compute-heavy workloads, where the relationship between where data lives and where computing happens is becoming increasingly important.

Related: A $37 million Meta scheme and five deputies earn a 'Godfather' 78 months in prison

Three customers, three very different problems

The first cohort was chosen to stress-test Shelby against distinct data and compute patterns rather than a single use case.

Teepin is building enterprise AI infrastructure around proprietary data and open-source models. Its integration with Shelby starts at the data layer, creating a path from storage into broader AI and compute services over time. 

Pictor Network is building infrastructure for distributed 3D rendering, where GPU capacity is spread across multiple locations and data availability directly affects how efficiently that computing power can be used. 

PathPulse is building spatial intelligence from video captured in the real world, a pattern that generates large volumes of data in one place that may need to be processed and analyzed somewhere else entirely.

Together, the three give Shelby very different environments to work against: proprietary enterprise data, distributed GPU workloads and high-volume physical-world data.

Trending on TheStreet Roundtable:

• McDonald's quietly changes how 14,000 stores price your Big Mac
• America's most taxed state pauses its plan to tax each crypto transaction
• Analyst warns Bitcoin is flashing a 2023 warning sign

The common thread

Across all three, a shared problem emerges. Data and compute don't stay in one place, and the infrastructure connecting them needs to keep up as workloads shift between providers, regions and hardware types.

That's the gap Shelby is designed to fill. Rather than forcing teams to manually copy, stage and synchronize data every time the compute environment changes, Shelby aims to make the data layer flexible enough that teams can choose infrastructure based on the workload rather than being locked in by where the data already sits.

A co-build, not a finished product

Aptos Labs is framing private beta explicitly as a co-build rather than a launch. 

Each customer engagement starts with the specific workload, where data lives, how compute is being used, which tools the application depends on, and the goal is to identify which requirements repeat across customers, which are unique and where the product roadmap should go next.

The company said it will share more detail on each customer's usage, data volumes and measurable results as the engagements progress. 

Shelby sits alongside Aptos' broader infrastructure push, which includes the recently announced MonoMove execution engine upgrade offering up to 55x faster smart contract execution and the Decibel exchange for onchain trading.

Related: Aptos unveils MonoMove, a 55x faster engine for onchain markets