Io.net
Decentralised network for GPU compute
io.net aggregates underutilised graphics processors from independent sources into a distributed cloud. Launched in 2023, it offers machine learning engineers lower-cost access to hardware for training and inference tasks.
- Launched at Solana Hackathon in Feb 2023
- Aggregates GPUs from crypto miners and data centres
- Supports parallel training and batch inference
- Offers lower costs than centralised cloud providers
- Handles orchestration and fault tolerance automatically
What it is
io.net operates as a decentralised physical infrastructure network that pools computing power from various sources. It connects independent data centres, cryptocurrency mining operations and other hardware providers to create a unified resource pool. Machine learning engineers use this system to run workloads at a fraction of the cost charged by centralised cloud giants. The platform handles orchestration and scaling automatically, allowing users to focus on model development rather than infrastructure management.
How it came about
The project began during the Solana Hackathon and the Solana Austin Hacker House in February 2023. Its founders aimed to address the shortage of available graphics processing units in the market. By leveraging the Solana ecosystem, they built a network that treats compute capacity as a tradeable asset. The initiative grew from a hackathon prototype into an enterprise-grade service for distributed computing.
The compute
The network supports parallel training, batch inference and hyperparameter tuning across many devices. It uses distributed computing libraries to split jobs among multiple graphics cards, overcoming memory limits on single machines. The system manages fault tolerance and scheduling for Python-based workloads. Users can export model weights to a shared object store for efficient serving. Reinforcement learning tasks are also supported through open-source libraries integrated into the platform.
Where the GPUs are
Compute resources come from a wide range of locations rather than a single corporate data centre. The network aggregates hardware from independent facilities and crypto mining farms. This decentralised approach allows the system to draw on underutilised capacity globally. The documentation does not list specific city locations for these nodes, emphasising the distributed nature of the physical infrastructure instead.
Customers and deals
The primary users are machine learning engineers and teams needing scalable compute. These groups often face long wait times and high prices from traditional providers like AWS or Azure. io.net targets those who require flexible access to hardware without the constraints of standard cloud contracts. The platform is designed for general-purpose computation with a strong focus on artificial intelligence tasks.
Money and ownership
Specific financial details such as funding rounds or valuation are not provided in the available documentation. The business model relies on reducing costs for users by tapping into cheaper, idle hardware sources. This economic structure aims to make large-scale model training and inference more affordable for startups and researchers.
Its place in AI
io.net positions itself as a solution to the global shortage of AI computing power. It challenges the dominance of centralised hyperscalers by offering a market-driven alternative. The project contributes to the decentralised physical infrastructure movement, turning idle hardware into a productive resource for the AI industry.
Questions people ask about Io.net
What is io.net?
It is a decentralised network that pools graphics processors from independent sources like crypto miners. This allows engineers to access cheap, scalable compute for AI tasks.
How does it reduce costs?
By using underutilised hardware from various providers, it avoids the high overhead of centralised data centres. Users pay less than they would for equivalent capacity from major cloud firms.
What tasks can it run?
The platform supports distributed training, batch inference, hyperparameter tuning and reinforcement learning. It is optimized for Python workloads and parallel processing.
Where did it start?
The project originated at the Solana Hackathon and Solana Austin Hacker House in February 2023. It was built to solve GPU scarcity issues in the AI market.
Good reading and sources
There is no English Wikipedia article on Io.net yet, so the facts on this page come from the sources listed above: docs.io.net/docs/what-is-io-net. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryNeocloud wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.