✍️Problem Statement

The advent of blockchain technology, artificial intelligence (AI), and machine learning (ML) has necessitated a paradigm shift in computational resource management. The GPU network, known for its robust parallel processing capabilities, serves as a cornerstone for these advanced computational domains. However, the burgeoning demand within public cloud environments has spotlighted significant challenges in accessing distributed computing resources efficiently. The decentralized computing power platform projects like io.net address this issue by integrating GPUs from underutilized resources. These resources are incorporated into a Decentralized Physical Infrastructure Network (DePIN), enabling engineers to access a vast amount of computing power in an accessible, customizable, economically efficient, and easy-to-implement system.

Despite the significant potential of decentralized computing power platforms, these kind of platfoms' current utilisation rate is generally below expectations, often under 10%. This low utilization is because the platforms' users are mainly AI developers, not ordinary end-users, presenting a barrier to using such computing power and resulting in considerable resource wastage. At the same time, there is a substantial demand for distributed computing power in the decentralized world, such as running blockchain nodes worldwide for L1 (e.g., Near), participating in AI chain mining with AI models (e.g., Bittensor), and mining in ZK chains with ZK algorithms (e.g., Aleo). However, for average users, running and maintaining such algorithms poses a high barrier.