Idle supply
Data centers built more GPUs than they can sell. Most run under 40% utilization, waiting for orders.
PAN is the demand layer for AI compute - AI usage routed to the world's idle data centers.
Data centers built more GPUs than they can sell. Most run under 40% utilization, waiting for orders.
People use AI every day, but that demand stays locked inside closed apps. Data centers can't reach it, and users don't get paid for it.
There's no common way for agents and apps to buy inference across providers, or to settle payment between them.
People chat, generate, and run agents on PAN-powered apps. Every request is real usage, not vanity traffic.
Every prompt is an order.
Requests get routed to partner data centers with spare capacity. What goes on-chain is a usage receipt, not a marketing claim.
Only the receipt goes on-chain.
Smart contracts pay data centers per call and pay users for the demand they create. No invoices, no delays.
Agents pay. You earn.
Powered by A2A Protocol →Three steps. The apps are free. Settlement is automatic.
No sales team. No idle racks. Paid-per-call, settled on-chain.
// Buy verified inference for 30 days, pay on delivery const res = await pan.a2a.request({ model: "qwen3-72b", window: "P30D", tokens: { n: 1e9, verified: true }, price: { max: 0.22, unit: "USDC/1M" }, proof: "receipt:onchain", }); res.settle(wallet); // → on-chain at N+2
Demand first. Protocol second.
Usage and revenue are on-chain receipts, not press releases. Audit us anytime.
Built around the Agent-to-Agent settlement primitive. Not adapted from legacy stripe rails.
Real usage before the token. Liquidity does not arrive — it is already paying.
They built the platforms. I'm building what platforms run on.
I didn't just build the community. I am the network.
Verification makes trust a protocol guarantee, not a promise.
Whether you're a builder, a data center, an AI app, or a network of agents — the layer is open.