AI Runtime runs two ways: the shared Gateway, and private or dedicated deployments inside a Home Cloud that belongs to you. A deployment is reviewed and activated with the AI Runtime team before it carries traffic — it is not a self-service switch.
What a deployment is
- Your Home Cloud. Google Cloud, AWS, Azure or an operator-approved private environment, chosen with your Agreement. The customer plane, ledger, secrets and evidence live there.
- A dedicated capacity plan. GPU class, expected request volume and scale behaviour are agreed before activation.
- The same contract as the Gateway. A deployment is reached through the same Pools, Workspace API keys, hard budgets and usage evidence. Your code does not change when the execution source changes.
What activation needs
- The Home Cloud and region where the deployment may run.
- The data-residency or network constraints that apply to your traffic.
- The expected request volume and latency target, so capacity is sized honestly.
- A signed Agreement covering the deployment and its support boundary.
Requesting a deployment
Use the contact form, or write to support@gridheap.com, and say which Home Cloud and region you
need. Include the Workspace that will call it. We will come back with the capacity plan, the cost
basis and the review questions for your environment.
What is not offered as self-service
Provisioning, resizing and deletion are operator actions today. AI Runtime does not publish a self-service deployment lifecycle until the customer-managed contract, artifact integrity, health evidence and support boundary are all complete.