# Compute overview

Canonical: https://savrn.com/cloud/docs/compute-overview

SAVRN Cloud · Coming soon. Public documentation and a browser simulation are available; connected services are not yet available.

Compute planning in SAVRN Cloud starts with the workload and the service offered to a project. This guide explains how supplier placement, resource limits and operating responsibilities fit behind that offer, using simulated allocation records rather than claims of available GPU stock.

**SAVRN Cloud · Coming soon · Public preview · Browser-local simulation · No live compute or payments**

## Explore the proposed supply layer

Open [Compute](https://savrn.com/cloud/console/#/compute), filter provider cards by region or source, then choose **Configure** to open the deployment form. Review model, GPU/region, hour ceiling, context, concurrency and route settings. Choose **Review allocation**, acknowledge the simulation boundary and choose **Create local deployment**. Inspect the resource in [Deployments](https://savrn.com/cloud/console/#/deployments).

These screens show the proposed SAVRN Cloud allocation workflow. They do not query supplier stock, request quota, create a virtual machine or verify a GPU. Every displayed resource and rate is a fixture unless explicitly linked as external evidence.

## Keep customer and supplier records separate

A customer chooses an eligible SAVRN Cloud service offer. Operators choose and qualify the supplier configuration behind that offer. The customer-facing record should preserve a stable project and request history when placement changes. Provider-specific lifecycle details belong behind an explicit adapter.

Provider names identify illustrative placement choices, not a supplier commitment or confirmed capacity. Any future rented or owned placement requires its own qualification and published service terms.

## Before connected services launch

Start with one supported configuration and a bounded operating window. Verify hardware, image, model, private access, resource lifetime, cleanup and actual cost. On-demand allocation, reserved capacity, multinode jobs and a marketplace are separate capabilities. Do not infer them from a compute catalog. Use [rented execution](https://savrn.com/cloud/docs/compute-rented-digitalocean) and [owned transition](https://savrn.com/cloud/docs/compute-owned-transition) to review their distinct acceptance requirements.