# Bread AI – decentralized AI inference using spare GPUs > Source: > Published: 2026-09-24 13:17:12+00:00 Simple. Powerful. Open Intelligence. An AI chat that runs on independent GPUs around the world. Ask anything for free. People who share their GPU earn BREAD for every answer they serve. Most AI today runs in a handful of hyperscale data centres: new buildings, new power lines, new hardware, all so a request can be answered somewhere far away. Meanwhile millions of capable graphics cards sit idle at home most of the day. BREAD AI makes better use of computing hardware that already exists. Every request is routed to an independent GPU that would otherwise be idle, the owner is paid in BREAD for the work, and no new hyperscale infrastructure has to be built for the workload to run. Not every AI job needs a data centre. Most of them just need a spare GPU and a fair way to pay for it. Type a prompt in BREAD AI. It becomes a job with a measured cost in BREAD, based on tokens. BREAD Console routes the job to an online GPU node. The model runs locally on that GPU and streams the answer back. Usage is settled on Solana Devnet: the node earns its share, the platform keeps its fee, every transaction is public. BREAD is what users spend for compute and what providers earn for supplying it. The price is set in cents and never surges: an answer costs the same whether the network is empty or slammed. What changes is the reward: when GPUs are scarce, hosts are paid extra. BREAD Node is a small desktop app. It detects your GPU, downloads the model once, connects outbound to the network and starts earning. No port forwarding, no accounts, no separate AI software to install. Version 0.1.0 (Devnet preview). Installer about 45 MB with the inference runtime included; the GPU backend for your card (up to 900 MB) and the model download on first run.