NuNet

What it does

NuNet is a protocol layer rather than a marketplace. Each participating device runs the Device Management Service, an open-source Go daemon that advertises capabilities (CPU, RAM, GPUGPUGraphics Processing Unit. Originally designed to render video game graphics, GPUs turned out to be exceptionally good at the massively parallel math that AI models need. Modern AI training and inference runs almost entirely on GPUs.Like a factory with 10,000 workers doing the same simple task in parallel, versus a CPU which is more like 10 workers each doing different complex tasks. AI training involves doing simple math a million times per second on a million numbers, which is exactly what the GPU factory is designed for.Read more →, geography, trust level) over a libp2p gossipsub mesh. The Logical Orchestration Layer matches incoming workloads to candidate devices via constraint satisfaction and dispatches the job to a containerised runtime. Each device and workload is an actor with a key pair; messages are cryptographically signed; the actor ID is the public key. Payment settles peer-to-peer in NTX on Ethereum, Cardano or BNB Smart Chain.

The project was incubated inside SingularityNET in 2018, incorporated as a standalone entity in August 2021, and shipped mainnet on 2 March 2026. Co-founders Dr Kabir Veitas (CEO and CTO), Dr Weaver D.R. Weinbaum, and Dr Ben Goertzel are all doxxed; Veitas runs day-to-day operations while Goertzel is now CEO of the Artificial Superintelligence Alliance. The dual-entity structure separates protocol stewardship from commercial operations: NuNet Foundation governs, NuNet Solutions builds.

One framing worth correcting up front. NuNet is sometimes described as the “remaining ASI Alliance compute leg” after Ocean Protocol’s October 2025 exit. That framing is wrong. The ASI Alliance lists three members: Fetch.ai, SingularityNET and CUDOS. NTX never migrated to FET. SingularityNET’s 2024 annual report describes NuNet as the “decentralised compute layer” supporting AGI infrastructure, not as a peer member. The SingularityNET Foundation retained 495,000 NTX from the 2021 spin-out (0.05% of supply) but has no governance control. Treat NuNet as infrastructure for SingularityNET, not as part of the ASI Alliance itself.

Value proposition

The pitch is architectural rather than commercial. Most compute marketplaces sit on top of a centralised orchestrator, take a cut, and sell decentralisation as a brand. NuNet inverts that. The orchestration layer is itself the open-source protocol. Anyone with the codebase can run a parallel matching engine. The Apache 2.0 licence on the Device Management Service and the NuNet Appliance makes that legally clean, and the libp2p mesh means there’s no central relay to gatekeep participation.

Multi-chain settlement is the second structural choice. NTX is deployed natively on Ethereum (about 631 million tokens), Cardano (about 369 million), and BNB Smart Chain. A consumer pays a provider in NTX on whichever chain both prefer. The protocol’s design explicitly accepts stablecoins and other settlement assets as alternatives, which solves a real interoperability problem and creates an obvious tension for token holders: if the protocol works equally well without NTX, what is NTX for?

The counter-narrative is just as direct. At 32 connected devices, NuNet is two orders of magnitude smaller than Akash, io.net, Render or Aethir. The Auki/Posemesh and Serverista partnerships are reciprocally confirmed but produce no public workload volume or revenue. The dashboard at dashboard.orgs.nunet.network returns an HTTP 307 redirect on plain fetch, so the only quantitative network state the project publishes is the 32-device figure on the homepage itself. There’s an open question whether NuNet’s honesty on scale is a feature (no inflated metrics to discount) or a constraint (no scale to underwrite a revenue thesis). Right now, both are true.

Where NuNet fits on the compute shelf depends on what you weight. Against io.net’s 327,000 registered GPUs and 6,720 daily verified active GPUs from the 2024 Sybil hangover, NuNet’s 32 reported devices are smaller, but they aren’t padding a marketing claim. Against Aethir’s $127.8M self-reported revenue with closed-source code, NuNet’s invisible revenue with open-source code is a different trade. Against Gensyn, which shipped a token in April 2026 with the training mainnet still pending, NuNet has the inverse profile: mainnet live, token deeply post-cycle.

Tokenomics

NTX is a fixed 1 billion supply with no inflation and no minting beyond the cap. As of mid-May 2026, 86.57% is already circulating, with the on-chain Ethereum supply read matching the CoinGecko figure. (CoinMarketCap publishes a different number; the on-chain read is authoritative.) Most dilution is already behind. The next unlock is 2.54 million NTX on 6 June 2026, which is 0.25% of supply and is genuinely negligible.

Distribution, per the project’s published token overview:

• R&D reserve: 21.00% (foundation-controlled release; no public schedule)
• Team & advisors: 16.25%
• Mining rewards: 15.25% (released to compute providers as workloads execute)
• Private round: 14.40%
• Community round: 10.00%
• Community rewards: 9.00%
• Backers: 6.00%
• Liquidity provision: 3.50%
• Seed round: 2.60%
• Launchpads: 2.00%

Insider allocation (team plus private plus backers plus seed) totals 39.25%. Community-allocated supply (community rounds plus mining rewards) totals 34.25%. The R&D reserve at 21% is the largest single category and it’s the structural overhang to watch, because the release cadence is foundation-controlled and not published.

The NTX Ethereum contract was audited by Hacken OÜ in September 2021 (Solidity 0.6.2, MIT licensed, verified on Etherscan). The report found no critical, high, medium or low severity issues. Hacken itself recommended additional independent audits and a public bug-bounty programme as part of the standard audit conclusion. Neither has been publicly executed in the four years and eight months between that audit and the March 2026 mainnet launch. The orchestration protocol (Device Management Service, Logical Orchestration Layer) deployed at mainnet has no published independent audit at all.

The Ethereum NTX contract has burn, mint, pause and role-based-access-control features per the verified bytecode. No documented burns. No buyback. No fee-share to holders. The “small transaction fees fund network operations and development” language on nunet.io doesn’t specify whether protocol fees accrue to NTX holders or to the NuNet Foundation treasury. There is no staking, no validator yield, no compute-provider yield-share, and no governance voting (yet). Compute providers earn NTX from completed jobs. Passive holders earn nothing.

Market context: NTX is down approximately 98.5% from its all-time high of $0.2293 set on 30 November 2021. The all-time low of $0.002906 was set on 5 May 2026, nine days before this review. An ATL set during the write of the review itself tells you sell pressure is still active rather than priced in. Listed on MEXC, XT.COM, Uniswap V3 (Ethereum) and Minswap (Cardano). No tier-1 CEX listing. See live data above for current pricing.

How to participate

Hold NTX. No staking, no yield, no governance voting yet. Passive participation is purely directional. Liquidity on tier-2 CEXes (MEXC, XT.COM) and DEXes (Uniswap V3 on Ethereum, Minswap on Cardano) means positions above retail size face material slippageSlippageThe difference between the expected price of a trade and the price you actually get when the trade executes. Slippage usually goes against the trader and gets worse with bigger trades or thinner markets.Like trying to buy 1000 bananas at the corner shop. The first few are at the marked price, but by the time you've bought them all you've moved the price up because there are no more bananas left at the original level. The shop has to restock at higher cost.Read more →.

Run a Device Provider node. Install the open-source NuNet Appliance (Ubuntu 24.04 base; MacOS and Windows supported). Minimum 2 CPU cores, 8 GB RAM, 100 GB storage, stable internet. The DMS advertises the device’s capability profile to the libp2p mesh and earns NTX peer-to-peer when matched workloads execute. The Appliance scales from a Raspberry Pi to a 64-GPU data-centre rack on the same actor framework. At 32-device network size, earnings expectations aren’t yet benchmarkable. Documentation at docs.nunet.io/appliance/about.

Use as a consumer. Submit compute jobs via the REST APIAPIApplication Programming Interface. A structured way for one piece of software to talk to another. In DeAI, APIs let applications request inference from a model without running the model themselves.Like a waiter in a restaurant. You don't walk into the kitchen and cook your own meal. You tell the waiter what you want, they tell the kitchen, the kitchen cooks it, and the waiter brings it back. The API is the waiter.Read more → at 127.0.0.1:9999. The protocol performs dynamic constraint satisfaction across location, latency, hardware type, cost and trust without manual provider selection. No public pricing or cost calculator is published.

Build on the SDK. Apache 2.0 codebase on gitlab.com/nunet; GitHub is a secondary mirror. UCAN capability-token authorisation model. REST API documented at docs.nunet.io/docs/developer-documentation. No NuNet-specific grants programme; SingularityNET’s Deep Funding accepts adjacent proposals.

Governance. Currently off-chain via the NuNet Foundation plus a Technical Council. Phase 2 of the governance roadmap introduces liquid-democracy voting weighted by NTX holdings, with the option to delegate to another holder or to an AI agent. Phase 3 targets development outsourcing. Phase 2 is roadmap, not deployed; mainnet shipped without on-chain voting.

Honest assessment

What works

The codebase is genuinely open. Apache 2.0 across the Device Management Service, the NuNet Appliance and the dev-tools. MIT on the Ethereum NTX contract. The primary repository lives on gitlab.com/nunet (GitHub at github.com/nunet-io is a mirror with 5 followers, which understates real activity). Anyone can fork the matching engine and run a parallel instance. That’s the architectural promise of permissionlessness, and unlike io.net or Aethir where the orchestration code is closed-source, NuNet actually delivers it.

Multi-chain settlement is operational. NTX trades and settles natively on Ethereum (~631M tokens), Cardano (~369M), and BNB Smart Chain. The Ethereum contract is verified on Etherscan with a clean Hacken audit on the token layer. The Cardano integration activated on 19 January 2026, six weeks before mainnet, with Eternl wallet support and recognition in the Cardano Foundation’s January 2026 monthly update.

Two production partnerships have reciprocal confirmation. Auki Labs published a partnership announcement on auki.com confirming joint node-deployment work within the NuNet Appliance for the Posemesh spatial-computing layer. Serverista’s integration was covered by Bitget News and describes NuNet’s Logical Orchestration Layer being used for dynamic VM workload distribution across data-centre infrastructure. Two reciprocally-confirmed partners is more concrete evidence than most early-stage compute protocols ship with, even before any revenue figure attaches.

The supply curve is the cleanest on the compute shelf. Fixed 1B cap with 86.57% already circulating. The next unlock is 0.25%. Compared to io.net (37.7% circulating, vesting through 2028) or Aethir (41.6% circulating, vesting through 2028), NuNet has the cleaner curve.

What doesn’t work

The audit gap is the headline weakness. The September 2021 Hacken review covered the token contract only and predates the orchestration protocol that shipped at mainnet. That’s a four-year-and-eight-month window between the only published audit and a live network handling peer-to-peer payments and containerised workloads. Hacken’s own report concluded with a recommendation for additional audits and a bug-bounty programme. Neither has been publicly executed.

Governance is fully foundation-controlled. No on-chain voting, no proposal register, no treasury transparency. The dual-entity structure (NuNet Foundation governs, NuNet Solutions operates) at least separates protocol stewardship from commercial interest, which is more than io.net offers, but it doesn’t substitute for an actual decision-making mechanism that token holders participate in. The Phase 2 liquid-democracy roadmap is appealing on paper. Until it ships, governance is one foundation board.

The token-utility model has a structural cap that the project itself describes openly. The protocol’s design accommodates stablecoins and other settlement assets, so NTX is the native asset but not the necessary one. Most live jobs probably do default to NTX, but the project is on record that they don’t have to. That’s honest. It’s also a permanent ceiling on the “NTX captures protocol throughput” thesis.

Network state is opaque. The 32-connected-devices figure on the homepage is the only quantitative network state the project publishes, and the underlying dashboard at dashboard.orgs.nunet.network returns HTTP 307 on plain fetch. There’s no public revenue, no transaction count, no compute-hours-delivered figure, no settlement-volume breakdown by chain. Project Catalyst funding documentation references 5,000 to 10,000 NTX transactions as an MVP success target, which is an honest acknowledgement of where the volume bar sits. Compare to Akash’s public Messari quarterly reports.

The risk

The structural risk is the supply pressure from the R&D reserve. 21% of supply sits in a foundation-controlled bucket with no published release schedule. At a $3M market cap, even a modest treasury draw covers operating costs for a meaningful runway, but it also sells into thin liquidity. The all-time low set nine days before this review (5 May 2026) suggests that pressure may already be playing out.

The narrative risk is dependency on SingularityNET, which is itself part of the ASI Alliance. The alliance’s governance has already been tested once: Ocean Protocol exited in October 2025 in a dispute settled with 286 million FET returned. NuNet’s separate token status (NTX, not migrated to FET) insulates it from analogous merger disputes. But if SingularityNET were to depart the ASI Alliance or wind down its NuNet integration, the most-cited consumer of NuNet infrastructure would be gone. Read that as a structural risk on the SingularityNET dependency; the codebase itself would survive.

The execution risk is timing. Mainnet shipped on 2 March 2026. Two and a half months in, the network has 32 reported devices, two reciprocally confirmed production partners, no public revenue, no governance voting, no working burn or fee-share, and a token at all-time-low. None of those gaps are technically blocked. All of them could be closed by the foundation. Whether they get closed before the supply overhang erodes patience is the question.

My position

Fact: NuNet ships an Apache 2.0 compute orchestration protocol with multi-chain NTX settlement, two reciprocally announced production partnerships, and a published 32-device network state.

Take: That’s the architecturally honest version of what most of the compute shelf is pretending to be. But “honest small” without a working burn, on-chain governance, or independent orchestration audit reads as a perpetual “show me” until the foundation closes those three gaps. At $3M market cap with the all-time low set nine days ago, the option premium is cheap; the option’s exercise still requires the foundation to execute.

I don’t hold NTX. If the foundation ships an audit of the orchestration protocol, publishes a verifiable network dashboard with monthly settlement volumes, and turns the dormant burn capability into an actual fee-funded burn, the position would be reconsiderable. None of those are speculation; they’re things the foundation has the codebase, the team and the runway to do.

Freedom Score: 47/100

NuNet scores 47/100 (D grade). Full methodology at Freedom Score Methodology.

Infrastructure decentralisation (10/20): Device onboarding is permissionless: anyone can install the open-source NuNet Appliance and join the libp2p mesh. Settlement runs across Ethereum, Cardano and BSC, so no single chain is a single point of failure. The Logical Orchestration Layer (matching engine) is a single foundation-operated reference implementation, however, and the network’s verified size of 32 devices is too small to demonstrate geographic or organisational distribution. Apache 2.0 across the DMS means anyone could fork the matching engine, but no alternative implementation exists.

Governance decentralisation (5/20): All current decisions sit with the NuNet Foundation and Technical Council. No on-chain voting, no public proposal register, no treasury transparency. The Phase 2 liquid-democracy model is on the roadmap but not deployed; mainnet shipped on 2 March 2026 without it. Above fully-team-controlled DePIN peers because the dual-entity structure (Foundation governance plus Solutions operations) at least separates protocol stewardship from commercial interest.

Token distribution fairness (6/15): Insider allocation totals 39.25% (team and advisors 16.25%, private 14.40%, backers 6%, seed 2.60%). R&D reserve at 21% is unusually large and represents future foundation-controlled release with no public schedule. Community-allocated supply (community round 10% plus community rewards 9% plus mining rewards 15.25%) totals 34.25%, which is reasonable for a DePINDePINDecentralised Physical Infrastructure Networks. Protocols that use token incentives to coordinate real-world physical infrastructure like GPU compute, wireless networks, storage, mapping sensors, or bandwidth.Like crowd-sourced ride-sharing but for physical hardware. Uber incentivises drivers with dollars. DePIN incentivises hardware operators with tokens. The network grows because individuals choose to contribute capacity in exchange for rewards.Read more →. The token launched via IDOIDOInitial DEX Offering. A token launch mechanism where the initial sale happens directly on a decentralised exchange, with built-in liquidity pool funding. IDOs replaced ICOs as the standard launch model for many projects after 2018.Like opening a new shop directly inside an existing market hall instead of running a separate sale event. The shop is open, the market provides foot traffic, and customers can walk in and trade from day one.Read more → and launchpad rounds rather than fair launch.

Censorship resistance (9/15): Apache 2.0 codebase means the DMS can be forked and run by anyone. The libp2p P2P mesh has no central relay. Multi-chain settlement (Ethereum, Cardano, BSC) means a single chain censoring NTX transactions still leaves alternatives. No KYC required for device participation. The Foundation could choose to disable its reference matching engine, but the protocol’s open-source nature limits the blast radius of any single foundation action.

Data sovereignty (7/15): Workloads execute locally on provider devices in containers; no centralised aggregation of model weights or inputs. Cryptographic signing of every actor message provides integrity guarantees. There is no confidential-computeConfidential ComputeHardware-enforced computation where data and code are encrypted in memory and only the authorised application can access them. The machine's operator cannot read what the application is doing even though they own the machine.Like renting space in a bank vault. The bank owns the building and runs the security, but what you put in the vault is invisible even to the bank staff. Only you have the key.Read more → hardware support (no Intel TDX, SGXSGXIntel Software Guard Extensions. The first widely-deployed TEE technology, introduced in 2015. SGX creates encrypted memory regions (enclaves) where code and data are protected from the operating system and the machine's owner.Like a safe deposit box at a bank. The bank owns the safe room and can see who comes in and out, but they can't see what's inside the boxes. SGX gives applications a private box on a shared computer.Read more → or NVIDIA H100/H200 confidential containers as offered by io.net and Phala), so sensitive workloads are limited by what providers’ standard environments support. Metadata (job matching, payment) flows through the DMS reference implementation which is foundation-operated.

Open source and transparency (10/15): Device Management Service, NuNet Appliance and dev-tools are Apache 2.0 on gitlab.com/nunet. NTX Ethereum contract verified on Etherscan, MIT licensed. Whitepaper 2.0 published. The Hacken token-contract audit (September 2021) is publicly accessible. The four-year-eight-month audit gap between the only published audit and the 2 March 2026 mainnet launch is the headline weakness; the orchestration protocol deployed at mainnet has no published independent audit. Hacken itself recommended additional audits and a bug-bounty programme; neither was publicly executed.

Path to improvement

Three changes would materially raise NuNet’s Freedom Score:

1. Audit the orchestration protocol. Commission a recognised firm (Trail of Bits, OpenZeppelin, Halborn or equivalent) to audit the Device Management Service and the Logical Orchestration Layer. The September 2021 Hacken token-contract review is not sufficient cover for a live mainnet handling peer-to-peer payments. Hacken’s own recommendation for additional audits has sat unactioned for four-plus years.
2. Ship Phase 2 governance. Move the liquid-democracy voting mechanism from roadmap to live, with a public proposal register, on-chain treasury operations and transparent R&D reserve release scheduling. The dual-entity structure is a start; on-chain voting is the proof.
3. Publish a verifiable network dashboard. The 32-device figure on the homepage should be cross-referenceable against a live, externally accessible network state page showing active devices by region, settled jobs over time, fee revenue by chain, and burn or treasury accumulation. Akash publishes quarterly Messari reports; NuNet currently publishes a homepage number.

Returns Score: 33/100

NTX scores 33/100 (F grade). Full methodology at Returns Score Methodology.

Token utility (9/20): NTX is the native settlement asset across Ethereum, Cardano and BSC with a real network-access requirement for consumers and providers. The protocol’s blockchain-agnostic design accommodates stablecoins and other settlement assets, which caps but does not eliminate NTX-specific utility because most live jobs default to NTX. No staking yield, no governance voting yet, no fee discount for NTX-denominated jobs. Compute providers earn NTX from completed jobs; passive holders earn nothing. Below io.net (12/20) and Aethir (12/20), which have payment plus staking utility layers.

Value accrual (4/20): Burn capability exists in the Ethereum NTX contract per the verified Etherscan bytecode and protocol fees are designed to flow through, but no public burn cadence, no documented burns, no buyback programme and no fee-share to holders. The “small transaction fees fund network operations” language on nunet.io doesn’t specify whether fees accrue to NTX holders or to the NuNet Foundation treasury. Marginal credit for a designed-but-not-fired mechanism versus zero infrastructure.

Supply dynamics (12/20): Fixed 1B cap, no inflation, 86.57% already circulating. Most dilution is already behind. Next unlock is 0.25% of supply in June 2026 and is genuinely negligible. Genuinely better than the compute-shelf peers where most dilution remains: io.net 37.7% circulating, Aethir 41.6% circulating, Gensyn pre-cliff with 54.6% insider allocation. R&D reserve at 21% is the largest structural overhang because release cadence is foundation-controlled and not publicly scheduled.

Revenue sustainability (4/25): Mainnet 2.5 months old. No published fee or transaction-volume figure. Production workloads named (Auki/Posemesh, Serverista, HOMEPUTE369, NuNet Native) and two are reciprocally announced from the partner side, which is more credit than self-claim alone. Project Catalyst documentation references 5,000 to 10,000 NTX transactions as an MVP target. No DeFiLlama or Token Terminal listing for fees. Similar to Gensyn (2/25, prediction-market fees nominal but real); well below Aethir (17/25, $127M self-reported) and io.net (15/25, $20M self-reported).

LiquidityLiquidityHow easily a token can be bought or sold without moving the price. High liquidity means you can enter or exit large positions quickly at the quoted price. Low liquidity means even small trades can swing the market.Like the difference between selling a house and selling a share of Apple stock. The house might be worth more on paper, but finding a buyer at that price takes weeks. The Apple share converts to cash in one click.Read more → and access (4/15): Approximately $3M market cap with thin daily volume per CoinGecko on the research date. Down 98.5% from the November 2021 ATH; ATL set 5 May 2026 (nine days before this review), indicating active sell pressure rather than a stable post-cycle base. Listed on MEXC and XT.COM (mid-tier CEXes) plus Uniswap V3 on Ethereum and Minswap on Cardano. No tier-1 CEX (Binance, Coinbase, Kraken) listing. Positions above $10k face material slippage on most venues.

Path to improvement

Three changes would materially raise NTX’s Returns Score:

1. Turn on the burn. The Ethereum contract has burn capability. Implement a publicly auditable monthly burn funded by protocol fees, with on-chain proof of every burn. This is the single most powerful signal that NuNet’s revenue flows to NTX holders rather than to the foundation treasury.
2. Publish revenue data. Settlement transactions on Ethereum and Cardano are on-chain; a Messari-style quarterly report listing settlement volume by chain, fees collected, and net-of-emission yield to providers would let the market price the network on revenue rather than on hope.
3. Earn a tier-1 CEX listing. MEXC and XT.COM are not the same liquidity tier as Binance or Coinbase. Listings follow revenue, audits and governance maturity. The other two paths-to-improvement above are prerequisites; the listing is the trailing indicator.

Score change log