# World Compute Full Agent Context

> Private GPU capacity for model training, inference, and evals without moving data outside the customer's network.

## Positioning

World Compute provides private AI infrastructure for teams that need GPU capacity without moving model data outside their network. The public tagline is "We sell supercomputers."

## Pages
- [Home](https://raidzero.com/): Product overview for private AI infrastructure and GPU capacity.
- [Agent view](https://raidzero.com/agents): Text-first World Compute overview with FAQs and machine-readable context.
- [Tutorial playground](https://raidzero.com/playground): Hands-on path for requesting GPU capacity and deploying model workloads.

## Agent Resources
- [llms.txt](https://raidzero.com/llms.txt): Curated entrypoint for AI agents and LLM crawlers.
- [llms-full.txt](https://raidzero.com/llms-full.txt): Expanded single-file context for larger agent windows.

## Compute Facts
- Private GPU capacity for training, inference, fine-tuning, and evals.
- Supported GPU classes include NVIDIA H100, H200, B200, and AMD MI300X.
- Cluster interconnects include NVLink and InfiniBand where the workload requires it.
- Workloads are designed to run inside the customer's network boundary.
- Billing is metered by the hour.

## Frequently Asked Questions
### What is World Compute?

GPU compute, metered by the hour.

### What GPUs can I rent?

NVIDIA H100, H200, B200, and AMD MI300X, connected over NVLink and InfiniBand.

### How does billing work?

Metered by the hour.

## Cloudflare Markdown for Agents

When Cloudflare Markdown for Agents is enabled for the zone, AI clients can request HTML pages with:

```sh
curl https://raidzero.com/ -H "Accept: text/markdown"
```

The origin keeps semantic HTML, metadata, JSON-LD, and these Markdown entrypoints available so Cloudflare can convert pages cleanly at the edge.

## Contact

- Website: https://raidzero.com
- Email: contact@raidzero.com