NuNet is a peer-to-peer protocol for discovering, orchestrating, and settling compute across distributed infrastructure. We connect fragmented computing resources — GPUs, servers, edge devices, IoT hardware, data centers — into a scalable, self-organizing network that anyone can join and use.
Computing is everywhere, but it doesn't work together. GPUs sit idle in data centers. Edge devices have unused capacity. Organizations can't share resources across boundaries.
Centralized cloud infrastructure fails when AI moves into the physical world — where latency matters, connectivity is intermittent, and costs compound at scale.
The internet solved this for information: independent networks connected through shared protocols so data flows end-to-end as one system. NuNet does the same for compute. Different compute pools stay autonomous, but become discoverable. Workloads route across them automatically, securely, and cost-effectively.
GPUs sit unused in data centers. Edge devices have spare capacity. There's no protocol to coordinate them.
Robots, energy systems, and IoT devices need sub-second responses. Round-tripping to a remote data center adds latency and creates single points of failure.
Organizations can't easily share or trade compute across ownership boundaries. NuNet makes resources from different owners securely poolable.
NuNet is not a cloud provider. Not a GPU marketplace. Not an edge device. Not a robotics OS.
NuNet is a compute orchestration and economic coordination layer — an open-source protocol (Apache 2.0) that sits between applications that need compute and the infrastructure that provides it.
Finds available compute resources across any infrastructure — GPUs, CPUs, edge devices, data centers — automatically.
Pairs workloads to resources based on processing power, location, latency, cost, and real-time availability.
Manages execution across dynamically changing sets of devices and owners. Workloads can move while running.
Every transaction settled peer-to-peer using NTX with cryptographically signed contracts. No intermediary.
The result: software finds compute autonomously. Not users browsing a marketplace. Not operators manually provisioning. The workload itself locates and secures the resources it needs.
Every architectural decision in NuNet traces back to one of these six principles. They're not aspirations — they're constraints we design to.
No central server dictates resource allocation. Each participant operates autonomously. The protocol is open-source (Apache 2.0) and vendor-neutral.
Cryptographically signed contracts between participants. Zero-trust architecture designed for machines that don't know each other collaborating on shared tasks.
The decentralized, self-organized architecture provides redundancy and fail-safe workflow design. No single point of failure. Workloads move between resources dynamically.
From a single sensor to a full data center. NuNet supports workflows that scale across any number of devices, owners, and locations.
Traditional cloud charges for maintaining underused infrastructure. NuNet mobilizes idle resources across the network, driving costs down through genuine utilization.
Designed by evolutionary scientists to assimilate new technologies as they emerge — new protocols, new hardware, new AI frameworks — without depending on centralized updates.
By 2029, the number of computational entities will be six times larger than the entire human population. Most of these devices spend hours every day idle. If coordinated during downtime alone, they would unleash a combined computational force of approximately 3×10²¹ FLOPS — enough to train 10,000 AI models per year, each comparable to today's most advanced large language models.
This isn't a hypothetical. The compute exists. It's just fragmented, disconnected, and underused. NuNet is the protocol that connects it.
NuNet was incubated as a spinoff of SingularityNET via the X-Lab Accelerator programme in 2018, with the mission of building a universal compute orchestration layer.
Today the team is 15 people, globally distributed from Sydney to Europe — 8 in core development — and the team that shipped Network Live.
Stewards the open-source protocol, maintains the network, and supports the community. Governed by a Council of 3–6 members and an Executive Fellowship. Incorporating in Switzerland.
Open Source · Apache 2.0Commercial arm building products and services on top of the open-source protocol. Open-core business model: the protocol is free and open, commercial applications are built on top.
Open-core modelDecentralized machine perception for AR, robotics, and spatial AI. NuNet deploys and orchestrates perception nodes across environments.
European smart housing project. Energy AI agents deployed inside homes and buildings, targeting ~30% electricity cost savings.
Using NuNet to provision VMs across distributed infrastructure. Turning fragmented capacity into a unified, addressable compute pool.
Network Live launched March 2, 2026 — delivering on the original whitepaper commitment and transitioning NuNet from R&D to adoption phase.
Install the NuNet Appliance and connect to the live network today
Onboard compute resources — CPU, RAM, disk, GPU
Join organizations and deploy workloads on distributed infrastructure
Pay and settle in NTX — peer-to-peer, no intermediary
Original project whitepaper outlining the protocol vision and architecture
Read → 📚Technical docs, guides, and API reference for builders and node operators
Read → ⌨️Open-source codebase. Apache 2.0. Contributions welcome.
Explore → 📡Monthly progress reports, roadmap updates, and what the core team is shipping
Read →