NCOMP · Token live on Robinhood Chain
Compute
Belongs to
the people.
Project trading fees are intended to fund open-model compute. First, prove a hosted pilot. Then buy hardware as demand justifies it. Inference and holder access are not live yet.
Don’t nationalize the labs. Nationalize compute.
Local AI is not optional anymore. Own the machines.
Why it has to be ours
Control the deployment. A provider can retire a model or change an API. Hosting open weights gives the operator more control over what keeps running.
Choose the model. Open weights let operators evaluate alternatives and configure their own deployments, subject to model licenses and applicable law.
Own capacity. Renting an endpoint means accepting its pricing and limits. Owning hardware adds control, along with power, maintenance, and operating costs.
Open weights need metal. Downloadable models still need memory, GPUs, and someone to keep the machines running.
Build shared infrastructure. NCOMP proposes remotely hosted, pooled GPUs. That is not the same as offline AI on your own computer; service terms and capacity still matter.
Compute is the new electricity. We should own more of the plant.
Why now
Access is not ownership.
A provider can change a price, a release, or an endpoint. Infrastructure ownership offers a different starting point.
Open models need somewhere to run.
Model availability is only half the problem. Affordable, reliable serving capacity is the other half.
Own the metal.
The proposal is simple: use trading fees to build a hardware reserve, then make that capacity useful. Publish purchases and operating results as they happen.
3%
planned standard trading fee
40%
proposed holder revenue share
100%
planned public curve launch
0
buybacks
Reserve status
Reserve cash · NVDA
—
Not published
Reserve cash · USDG
—
Not published
Hardware · book
—
No hardware acquired
GPUs online
—
Not live
Distributed · USDG
—
No distributions
The proposed loop
- 01 A standard buy or sell pays 3%: 1% base fee plus 2% creator tax.
- 02 The fee model allocates 90% to the Reserve and 10% to Pons — 2.7% and 0.3% of the trade, including the assumed base-fee creator share.
- 03 Project fees accrue in NVDA for the creator wallet to claim from Pons.
- 04 Hardware funding is then a manual, wallet-controlled step. The plan is to publish GPU purchases and hosting costs; fees do not automatically enter a FeeVault or Treasury.
- 05 The GPUs would serve open models: paid inference in USDG and a holder-access tier.
- 06 A draft 60 / 40 inference-revenue split proposes reinvestment and a holder share. Costs, eligibility, and payout terms remain to be finalized. Trading fees are excluded from distributions.
The token is live; the inference service and revenue-sharing design are not. The fee split above remains a model pending reconciliation with actual Pons receipts. Initial anti-snipe charges can exceed the standard fee.
For holders
Proposed access tiers: 100,000 NCOMP for basic; 1,000,000 for full. Quotas, model availability, and balance-check rules are not final.
The draft proposes a 40% holder share of inference revenue in USDG through balance snapshots and a Distributor. No inference service or distributions are live.
The design does not require staking or a lockup. Token ownership alone does not guarantee service, income, or ownership of hardware.
The hardware
| Node | GPUs | Location | Acquired |
|---|
No hardware acquired yet. The next milestone is a hosted pilot with measured usage and operating costs, before committing to owned hardware. Any rented capacity will be identified separately from owned assets. A legal entity and colocation arrangement are not established.
Tokenomics
NCOMP is live on Pons, paired with NVDA on Robinhood Chain. The verified token supply is 1,000,000,000 NCOMP. Use the contract address below to verify the asset before trading.
Planned standard fee: 3% on buys and sells, with a modeled 90% / 10% split between the Reserve and Pons.
That models 2.7% of the trade to the Reserve and 0.3% to the venue. The project/venue allocation must be reconciled against actual Pons fee receipts; it is not a published operating result.
Trading fees are intended for hardware, not buybacks or holder distributions. The creator wallet claims and controls these funds. Future Treasury contracts would also leave spending under owner control; payment memos do not enforce this policy.
Addresses
| Contract | Address | |
|---|---|---|
| Treasury | Not deployed · optional infrastructure |
Blockscout |
| FeeVault | Not used for creator-wallet claims |
Blockscout |
| Distributor | Not deployed |
Blockscout |
| NCOMP | Published at launch |
Blockscout |
| NVDA pair | 0xd0601CE157Db5bdC3162BbaC2a2C8aF5320D9EEC |
Blockscout |
Questions
What backs NCOMP?
The NCOMP token is live, but no hardware has been acquired and no inference service or revenue distributions are live. Optional Treasury, FeeVault, Distributor, and ComputeIndex contracts are not deployed. A token is not a redemption right or legal title to hardware.
How do holders earn?
The draft proposes a 40% share of inference revenue in USDG through balance snapshots and a Merkle Distributor. Costs, eligibility, and payment terms remain open. No earnings or distributions are live or guaranteed.
What are the free tier thresholds?
The draft proposes 100,000 NCOMP for basic and 1,000,000 for full access, without staking. The service is not live; quotas and final access terms will be published before it opens.
Where do the fees go?
The intended standard fee is 3%, modeled as 90% to the Reserve and 10% to Pons, including the assumed base-fee creator share. Project fees are intended for hardware. Initial anti-snipe charges and final on-chain routing must be checked separately.
How do I verify?
Read the chain. Robinhood Chain has chain id 4663; the explorer is robinhoodchain.blockscout.com. The addresses above are published with Blockscout links.