GPN — One-Pager

What it is. GPN is a GPU-native network where every node is both a miner and an AI provider. One install, one card: mining runs whenever the GPU is free, and an AI job pre-empts it, runs on the whole GPU, then hands it back.

The problem. A GPU owner has two ways to earn — mine a coin, or rent the card for AI work — and must pick one. Mine and the coin dips → you lose. Rent and nobody books you → $0. The card can't move to where the money is.

The fix. GPN removes the choice: the node mines by default and yields the GPU to each AI job, so the same card earns from both markets. The chain does coordination + escrow + settlement; the heavy compute runs off-chain.

How it works

• Mining — GPU proof-of-work; the node that produces a block keeps that block's reward (Bitcoin-style).
• Serving AI — LLM inference, image/video generation, ML inference, fine-tuning — priced per token against real cost, escrowed trust-minimally.
• The switch — per job, not per market cycle: an incoming job pauses mining, frees VRAM, runs on the whole GPU, then mining resumes. One GPU runs exactly one job at a time.
• Routing — buyers hit one endpoint; the network picks the lowest-latency, least-loaded verified GPU. Providers are discovered on-chain.

Why it's different

• Both markets, one machine — no incumbent (NiceHash, Vast, RunPod, Render, Akash, io.net) is both a native PoW miner and an AI marketplace on the same card.
• Trust-minimal — on-chain escrow, optional payment channels (no per-job block delay), opaque/private on-chain values.
• GPU-only, verifiably — proof-of-work + hardware attestation + a compute challenge; a CPU/iGPU can't earn.

Token (GPN)

• Fair launch via mining — no ICO, no hidden allocation. 12 decimals.
• Halving every ~2 years (4,204,800 blocks); block reward 50 GPN @ 15 s.
• Hard supply cap ≈ 520 M GPN — 420 M emitted over the halving schedule plus 100 M allocated to the treasury.
• Disclosed, vested treasury — 100 M GPN held by a keyless pallet account; 70 M unlocked at genesis, 30 M vesting over 10 years; spendable only by a 13-of-25 governance roster.
• Token value derives from real AI-compute demand — not just mining inflation.

Privacy. A shielded pool supports private transfers and withdrawals (zero-knowledge spends; nullifier-based double-spend protection).

Where it is. Live on a public testnet: PoW chain, per-node rewards, the full AI marketplace (LLM/image/video/ML/fine-tune), payment channels, shielded spends, on-chain provider discovery, peer discovery, committee governance. Next: the public CRS ceremony, mainnet, and a cross-chain bridge.

One line: A GPU should not have to choose.

gpncoin.ai · whitepaper + docs at https://gpncoin.ai/docs/