AI and agent infrastructure
CloudSpinUp runs a metered agent and memory plane on its own infrastructure. It is real, it is instrumented, and it is deliberately not self-serve. Access is granted per tenant after an assessment, on agreed ceilings, rather than sold as an instant tier.
State of the lane
- Access
- By assessment
- Scope, data boundary, and fair-use ceilings are agreed before a tenant is enabled.
- Entitlements
- Enforced per tenant
- Job, token, embedding, storage, and compute ceilings resolve against the tenant's access plan at call time.
- Metering
- Live
- Jobs and embeddings accrue against the tenant meter and reconcile into chargeback.
- Self-serve provisioning
- Not enabled
- GPU scheduling is operator-run. There is no automated allocation path, so there is no deploy button.
Published AI plans
source: catalog-serviceReading the catalog…
Why there is no deploy button
GPU allocation on the AI node is not automated. An order placed here would record billing intent and hand back an endpoint that nothing is listening on. A button that cannot deliver is worse than an honest request, so this surface asks for the request instead.
The metered agent and memory plane is a separate matter: it is already serving tenants, already enforcing entitlements, and already accruing usage. What it does not have is a self-serve front door, and this page will not pretend otherwise.
Request access
Tell us the workload, the data boundary, and the volume you expect. We come back with scope, ceilings, and price.
Looking for something you can order today? Compute is the lane with a delivery path behind it.