Skip to content

What is AIGr.id?

AIGr.id (pronounced "AI Grid") is an open network that lets many independent organizations and individuals pool their computers and AI models and use them together, as if they were one large AI platform. No single company owns or controls it.

The software that runs the network is OpenOS.AI, also called AIOS. AIOS is a distributed operating system for AI. It is 100% open source and community-driven.

In one sentence

AIGr.id is to AI what the internet is to information: many independently owned pieces, connected by open rules, that together become far more useful than any one of them alone.


The problem it addresses

Today, most advanced AI is built and run inside a few large, closed platforms. That creates four problems:

Problem What it means in practice
Centralized Building and running AI needs large budgets, large data centers and specialized teams, which shuts most of the world out.
Controlled A small number of companies decide what gets built, who can use it and on what terms.
Limited participation Only a small group can produce and distribute AI, so the benefits are spread unevenly.
Fragmented AI systems are siloed. There are no open standards for them to discover each other and work together.

AIGr.id takes a different approach. Instead of one big system, it connects many smaller ones and lets them cooperate.


How it works, in plain terms

flowchart LR
    A["People and organizations<br/>contribute computers"] --> N(("AIGr.id<br/>network"))
    B["Developers publish<br/>AI models and tools"] --> N
    C["Communities write<br/>rules (policies)"] --> N
    N --> D["Users and applications<br/>send AI tasks and get results"]
  1. Contribute compute. Anyone with spare hardware, from a single machine to a data center, can join it to the network. Machines are grouped into clusters, and clusters join a network.
  2. Publish AI. Developers package an AI model, or any piece of computing logic, as a reusable Block. Blocks run on the contributed hardware and grow or shrink automatically with demand.
  3. Compose. Blocks can be chained into workflows called vDAGs. For example, detect objects → track them → estimate their pose. Each step can run on a different cluster, in a different place.
  4. Govern. Every important decision is controlled by policies: small, readable Python programs. Policies decide who can join, where work runs, how it scales and who can use what. Each network, cluster and Block owner writes their own.
  5. Use. Applications send tasks to a Block or a vDAG through an inference endpoint and get results back.

You'll find these terms explained in more depth in Key concepts and the Glossary.


What makes it different

  • Open and shared. It is designed as global public infrastructure for AI, a digital commons built by many, for all.
  • Sovereign. Communities, companies and countries can own, control and govern their own part of the network without depending on an outside provider.
  • Plural and polycentric. Many models, many owners and many rulebooks can coexist and coordinate. There is no "one true model".
  • Composable. Small, reusable AI building blocks can be combined into larger systems, and one Block can serve many workflows at once.
  • Efficient. Several workloads can share one GPU, and resources are allocated by policies that the owners control.
  • Programmable rules. Governance is written in Python, so the rules are explicit, transparent and testable.

Who it is for

You are... AIGr.id lets you... Start here
A business or public organization Run your own private or sovereign AI grid on your own hardware, or join a shared one, while keeping control of data, models and rules. FAQ, then Architecture
An infrastructure owner Contribute clusters or individual machines and make spare GPUs and CPUs useful. Join a cluster, Join a node
A network operator Start a new network and define its membership and governance rules. Create a network, Network admin guide
An AI / ML developer Package models as Blocks, compose them into workflows and serve large models split across machines. Quickstart, Build
An application developer Call AI Blocks and workflows from your app over gRPC. Inference user guide
A researcher in governance, ethics or economics Design and test the policy layers that govern how AI is shared and used. Policies

Example uses

  • Internet of Intelligence. A global, open network where AI services discover and use each other.
  • Collective intelligence. Many specialized models, owned by different groups, working together on one task.
  • National sovereign AI grid. A country links its universities, data centers and public institutions into one governed AI infrastructure.
  • Enterprise sovereign AI grid. A company connects its offices, clouds and on-premise GPUs into one private grid with its own rules.
  • Community federated AI grid. A community pools its hardware and shares models under rules it agrees on.

Project status

Beta software

A variant of AIGr.id has run in production for close to a year at nearly 500,000 inferences per second on bare-metal, federated infrastructure, serving real-time vision AI workloads. The released version adds significant upgrades, including support for large language models. It is unit and integration tested but has not yet been validated at that scale or duration.

The project is therefore in beta and not recommended for production use yet. See the roadmap on the home page.


Where to go next