GPU Provider Reward

The GPU Provider Reward mechanism incentivizes the contribution of computing resources to maintain network stability and balance short-term rewards with the network's long-term growth.

The reward system has three key components:

Total Reward = Usage Fee + Computing Reward + Committed Compute Power Reward

Usage Fee

Final Usage Fee = Usage Fee x (1 - Network Fee%)

GPU Providers charge clients a Usage Fee based on the computing resources required for the task and the client's willingness to pay. Nimble Network will then collect a 5% Network Fee from the Usage Fee to cover network maintenance and development.

Computing Reward

Computing Reward = Reputation Score x Base Computing Reward

A GPU Provider's Reputation Score is determined by their performance record and trustworthiness. Higher-scoring providers will be prioritized for more AI tasks, encouraging them to maintain a strong professional reputation. GPU Providers can boost their Reputation Score by staking more $NIM tokens, consistently fulfilling AI tasks, and avoiding fraudulent activities.

The Base Computing Incentive is calculated based on the number of AI tasks completed, their complexity, and the GPU model used. An AI OrderBook system efficiently matches computing needs with available resources, ensuring optimal pricing and execution of AI tasks. This dynamic system allows for efficient resource allocation and adjusting fees and rewards based on real-time market conditions.

Committed Compute Power Reward

GPU Providers are rewarded by committing their computing power to the network. They need to regularly update their availability periods and pricing through a weekly Proof of Commitment. This helps ensure their hardware is always ready for use when required, thereby contributing to a stable and efficient infrastructure.

Rewards are determined by several factors, including the number of GPUs each provider contributes, the specific GPU models, the providers' availability and uptime, and the total number of GPUs available across the Nimble Network.