> ## Documentation Index
> Fetch the complete documentation index at: https://docs.b3iq.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

Own GPU machines, run your workloads, and choose whether idle capacity earns with B3IQ.

B3IQ is a neocloud that belongs to its customers. Buy a bare-metal GPU machine,
host it with B3IQ in Eugene, Oregon, or take delivery. Your account owns the
machine. Its gateway keys route only to machines on that account.

**Current service boundary**

Inference on your own machines has no per-request B3IQ charge. Optional Earn
  mode offers idle capacity by the hour and removes that machine from your
  inference routes. There is no shared public inference pool or third-party
  provider fallback. If your owned machines cannot serve a request, it fails.

## Start with your machine

**[Choose hardware](https://b3iq.org/machines)**

Compare current configurations, hosting and delivery options. Model guides
    help plan memory; they do not prove that a deployment is installed or ready.
  **[Use your account](https://b3iq.org/dashboard/nodes)**

Claim your provisioned machine, inspect its connection and runtime, and
    manage the models and keys your application uses.
  **[Call the gateway](/developers/hosted-gateway)**

Use OpenAI-compatible HTTP endpoints with an account-scoped gateway key.
    Discovery reflects your machines and that key's scope.
  **[Use local inference](/quickstart)**

Your provisioned host also supports local setup, local API keys and local
    OpenAI-compatible inference without managed enrollment.

## Choose what idle capacity does

Earn is optional. A machine in Earn mode is unavailable to your B3IQ inference
routes until you return it to Bare metal. Under Managed Hosting, the owner
receives 85% of rental revenue before hosting costs are deducted. Utilization,
market rates and costs vary; income is not guaranteed.

See [claiming and earnings](/operators/earnings) and the
[ownership terms](https://b3iq.org/ownership) for the distinction between
estimated earnings, accrued earnings and paid funds.

## Privacy and operational evidence

**[Private jobs](/developers/private-jobs)**

Send encrypted asynchronous work to a machine your account owns.
  **[Privacy boundary](/protocol/privacy)**

Ordinary gateway transport and encrypted private transport have different
    visibility and storage boundaries. Confidential execution remains planned.
  **[Check readiness](/operators/route-readiness)**

Inspect machine health, installed models and key restrictions; verify a
    small request with the same key your application uses.

## In the press

**[Fortune](https://fortune.com/2026/08/11/ai-startup-b3-has-found-a-compute-niche-with-a-rent-to-own-model-aimed-at-academics-clients-include-stanford-and-nyu-researchers/)**

Coverage of B3IQ's machine-ownership model for researchers. August 11, 2026.
  **[Pantera Capital](https://www.youtube.com/watch?v=9cKS6FH99Xo)**

Why AI Builders Need to Own Their Compute. Interview on the investor's channel.

## Historical surfaces

The public node explorer, speed leaderboard and inference-rate pages were
retired in the October 3, 2026 source change. Old links lead to current machine,
hardware and earnings-planning pages. Historical receipt and settlement records
are retained; they do not describe current per-request earnings or a public
serving network. [Current status](/operations/current-status) separates source
capabilities, historical pilot evidence and deployment acceptance.