Cost compounds with scale.
Cloud bills grow faster than the workloads that create them. Pricing is opaque, unpredictable, and structured to make exit expensive. The bigger you get, the more locked in you are.
NuNet is a peer-to-peer protocol that discovers, matches, and delivers compute across any infrastructure — GPUs, servers, edge devices, data centers — so workloads find the right resources automatically.
Build on NuNet →Foundations, alliances, and strategic partners NuNet works with on open infrastructure, decentralized compute, and AI. Live deployment partners are in the deployments section below.
AI is moving into the physical world. Into robots, buildings, vehicles, factories, and cities. These systems need compute that's local, fast, and coordinated across devices and locations. Traditional cloud can't deliver that.
Cloud bills grow faster than the workloads that create them. Pricing is opaque, unpredictable, and structured to make exit expensive. The bigger you get, the more locked in you are.
Robots, vehicles, and smart buildings need sub-second decisions. A round trip to a data center hundreds of miles away isn't fast enough. Physics doesn't bend for your region selection.
Millions of GPUs sit idle while demand surges elsewhere. Without a protocol to discover, match, and deliver compute across owners and locations, supply and demand never find each other.
What if compute could find workloads, instead of the other way around?
ContinueNuNet turns compute into a protocol. Devices join. Workloads find them. Payments settle. No marketplaces, no queues, no middlemen.
Install the NuNet Appliance on any machine. GPU rigs, edge servers, data center nodes, even a laptop. It registers your device on the network and makes its resources discoverable.
The orchestration layer matches workloads to available resources. Based on requirements like location, latency, GPU type, and cost. No manual provisioning. No marketplace browsing.
Every transaction settles peer-to-peer using NTX. Providers get paid directly. No invoices, no middlemen, no 30-day payment terms. Just compute delivered and settled.
Six architectural decisions that separate NuNet from marketplaces, queues, and rental platforms. Infrastructure that orchestrates itself.
Even rental layers shipping "smart deploy" still hand you a node list. NuNet’s orchestration is part of the protocol — workloads describe what they need, the network finds the match.
→ protocol-layer routingAllocations move across devices and locations as demand shifts, hosts fail, or a closer match becomes available. Stateless workloads migrate seamlessly; stateful workloads checkpoint and resume.
→ adaptive allocationZero-trust architecture with cryptographic verification at every layer. Compute from untrusted sources is isolated, verified, and auditable.
→ verifiable computeGPU-rental layers can’t run robots, sensors, or edge devices. NuNet runs the same actor framework on a Raspberry Pi, an on-prem server, or a 64-GPU rack — without per-tier rewrites.
→ any hardwarePool compute across organizations, geographies, and trust boundaries. Resources from different owners work together seamlessly.
→ federated poolsNTX settlement is built into the protocol — not bolted on. Every compute transaction has a native economic layer for pricing, payment, and incentives.
→ protocol-level settlementReal partners. Real workloads. Running on the network today.
AI agents managing energy consumption across smart buildings. Data is processed locally on edge devices instead of being sent to the cloud.
Decentralized machine perception for AR and robotics. Spatial anchoring and pose estimation computed across distributed edge nodes.
Automated VM provisioning and resource allocation across data center infrastructure, orchestrated through NuNet's protocol layer.
Autonomous AI agents that discover and provision their own compute resources. Scaling up, moving between providers, and settling payments without human intervention.
NTX is the utility token that makes decentralized compute orchestration work. Every interaction on the network flows through NTX. Utility only. No yield promises, no returns.
NTX is the access medium for the compute network. Providers and consumers use NTX to join the network and commit to service quality.
A small protocol fee in NTX is collected on each transaction to fund network operations, development, and governance.
Peer-to-peer payment for compute services. Providers are paid directly in NTX. No intermediaries, no invoicing delays.
While NTX is the native token, the protocol supports settlement in stablecoins and other assets, making it accessible to enterprises.
// UTILITY NOTICENTX is a utility token, not an investment product. Nothing on this site constitutes financial advice. Review the NTX tokenomics documentation for full details.
NuNet is a peer-to-peer protocol that discovers, orchestrates, and settles compute across distributed infrastructure. It connects GPUs, servers, edge devices, and data centers into a coordinated compute fabric. The protocol launched live on March 2, 2026, with research incubated since 2018.
NuNet Network Live launched on March 2, 2026, delivering the protocol described in the original NuNet whitepaper and refined in Whitepaper 2.0 (2021). The network runs real workloads today in energy (HOMEPUTE369 smart buildings), robotics (Auki / Posemesh), and data-center infrastructure (Serverista). Live metrics are published at dashboard.orgs.nunet.network.
NuNet settles compute transactions peer-to-peer in NTX on both Ethereum and Cardano. The protocol is blockchain-agnostic by design and supports multi-token settlement for compute payments. Orchestration fees settle in NTX, aligning coordination with the protocol's utility layer across both chains.
Most decentralized compute networks work like marketplaces: you browse available hardware, pick a node, and deploy manually. They're also GPU-focused. NuNet is different in four ways. (1) Orchestration, not browsing — describe what the workload needs (latency, location, hardware type) and the protocol finds the right compute automatically, no node picking. (2) Any hardware, not just GPUs — NuNet runs across GPUs, CPUs, edge devices, Raspberry Pis, and data center racks. (3) Multi-owner by design — built for workloads spanning different owners, sites, and trust boundaries, not a pool of identical GPUs under one operator. (4) Orchestration and payment combined — most networks settle after a job finishes; NuNet handles matching, verification, and multi-chain settlement as one unified protocol layer.
NuNet was incubated in 2018 inside SingularityNET by Dr. Kabir Veitas (CEO/CTO), Dr. Weaver D.R. Weinbaum, and Dr. Ben Goertzel. The company was incorporated in August 2021. The protocol is stewarded by the NuNet Foundation and commercialized through NuNet Solutions. Team size: 15 globally distributed members, 8 in core development.
NTX is the utility token that powers participation in the NuNet network. It has four utilities: staking to join the network, peer-to-peer payment between compute buyers and providers, orchestration fees paid to the protocol for workload matching, and multi-token settlement flexibility. Total supply: 1 billion NTX. Circulating: around 504 million.
Install the NuNet Appliance on your machine. It works on GPU servers, CPU-only servers, Raspberry Pi, and data-center racks. After installation, your hardware joins the network and becomes discoverable to workloads. You earn NTX for compute delivered. Full onboarding guide is in the documentation at docs.nunet.io/appliance/about.
Technical documentation is available at docs.nunet.io. The NuNet Whitepaper at docs.nunet.io/reference/white-paper describes the protocol's design, and the Yellow Paper at docs.nunet.io/reference/yellow-paper/platform-yellow-paper/ covers tokenomics and settlement. The open-source codebase is published on GitLab under Apache 2.0 license.
Builders, operators, and researchers shaping decentralized compute. Follow the protocol, join the conversations, or contribute to the code.
Protocol updates. Release notes. Ecosystem news as it happens.
FollowLive discussion with the team, developers, and contributors. Where the real work is debated.
Join serverCommunity chat. Announcements, questions, quick threads.
JoinTechnical walkthroughs, demos, and conference talks.
WatchCompany news, hiring, partner announcements, industry analysis.
FollowWhere the engineering lives. Source, issues, merge requests, CI pipelines, releases, and technical documentation for every NuNet repository. Apache 2.0 licensed and open to contributors.
Browse repositories