Apus Network

Apus Network is a decentralised GPU network built on Arweave and AO, started in 2023 to supply verifiable compute for AI training and inference. Model files are stored on Arweave with unique identifiers, an AO process receives inference requests and dispatches them to connected GPU nodes, and results plus logs are written back to permanent storage.

Trust comes from combining storage-based consensus with optimistic machine learning verification: programs, messages, models and logs all land on disk so anyone can recompute them, and GPUs are registered through decentralised physical device identifiers so compute comes from managed hardware. In June 2024 the AO ecosystem launched its AI on AO programme with Apus handling the compute scheduling layer that connects open models to onchain apps; after AO mainnet went live in February 2025, inference tasks settled through mainnet processes.