Payment and transaction fees within the io.net ecosystem, such as purchasing/providing underutilized GPU computing power, deploying GPU clusters.

Users can participate in network security and earn a share of idle GPU rewards by doing any of the following: Delegators stake $IO tokens on a node of their choice. Stake more $IO tokens directly to run on their own nodes.

IO tokens enable holders to participate in the governance of the IO Grants DAO by proposing, voting, and implementing funding proposals, which guide the allocation of funds and strategic direction of the GPU Internet Foundation.

Development companies that maintain the IOG network and build products and services on top of it

Also known as the Internet of GPUs or "IOG", the IOG Network is a decentralized physical infrastructure network consisting of independently operated, geographically distributed hardware nodes that provide permissionless access to on-demand computing.

Open source, developer-friendly SDK that enables builders to deploy products and services on the Internet of GPUs

io.net was created to build a globally accessible decentralized cloud computing network that provides affordable, flexible, and permissionless access to computing power.

With the rise of artificial intelligence, GPU computing power has increased significantly, while there is a severe shortage of GPUs in public clouds. This has caused problems such as long waiting times, limited choices, and high costs. By aggregating GPUs from underutilized sources such as independent data centers, idle crypto mining farms, and consumers, io.net can provide computing power at 90% lower cost than centralized cloud providers.

Batch Inference and Model Serving: Inference can be performed on incoming batches of data in parallel by exporting the trained model’s architecture and weights to shared object storage. io.net allows machine learning teams to build inference and model serving workflows across a distributed network of GPUs. Parallel Training: CPU/GPU memory limitations and sequential processing workflows present a huge bottleneck when training models on a single device. io.net leverages distributed computing libraries to orchestrate and batch training jobs so that they can be parallelized across many distributed devices using data and model parallelism. Parallel Hyperparameter Tuning: Hyperparameter tuning experiments are inherently parallel, and io.net leverages distributed computing libraries with advanced hyperparameter tuning capabilities to check for optimal results, optimize scheduling, and simply specify search patterns. Reinforcement Learning: io.net uses an open source reinforcement learning library that supports production-grade, highly distributed RL workloads with a set of simple APIs.