Provisioned capacity
Provisioned throughput, sized from every minute of use
UnitVane AI reads per-minute utilization for every provisioned deployment, recommends the size each one needs, and shows how much of what you deploy is covered by reservations, without buying anything until you approve it.
Right-sizing
Each deployment sized to the percentile you choose
Per-minute telemetry is rolled up to percentiles over a 30-day window. Each deployment gets a recommended size at your sizing percentile, with deployments that need at least 7 days of data before they are sized. Latency-sensitive deployments can be pinned to PTU; others are compared against pay-as-you-go.
- Per-minute data, not hourly averages
- Spillover-aware: routes with pay-as-you-go overflow size smaller
- Break-even volume, and how long a move off PTU stays cheaper
Reservation coverage
Coverage computed from what you deploy after right-sizing
Global, Data Zone and Regional capacity are separate pools and separate reservation products. UnitVane AI compares reservations in each pool with deployed PTUs after right-sizing, and flags capacity committed beyond peak deployment and capacity deployed steadily but not reserved.
- Pool-aware: Global, Data Zone and Regional
- Over-reserved and uncovered PTUs, per pool
- Renewal timeline aligned with current savings
Purchase automation
Reservations bought and renewed only with approval
Set a policy per pool: no automation, a monthly purchase, or renewal on expiration, at shared, subscription or management group scope. Recommendations only ever reduce cost, and every purchase waits in pending approvals for a named approver.
- Monthly, 1-year and 3-year terms
- Pending approvals with a full audit trail
- Pay-as-you-go advice routed to team review instead
# Reservation automation, per pool. Purchases always wait for approval.apiVersion: ai.unitvane.com/v1kind: ReservationPolicymetadata: name: data-zone-usspec: pool: datazone/us mode: renew-on-expiration # off | purchase-monthly | renew-on-expiration term: P1Y # P1M | P1Y | P3Y scope: shared # shared | subscription | management-group coverTo: deployed-after-rightsizing approval: required: true approvers: [finops-leads]Why AI spend needs an owner
Supported capacity
Provisioned capacity on every cloud
| Provider | Capacity unit | What UnitVane AI sizes |
|---|---|---|
| Azure OpenAI | PTU, Global, Data Zone and Regional | Deployment size, reservation coverage per pool, renewals |
| Amazon Bedrock | Provisioned throughput model units | Units per model, commitment term, on-demand comparison |
| Google Vertex AI | Provisioned throughput (GSUs) | Units per model and region, pay-as-you-go comparison |
| Self-hosted models on GPUs | GPU nodes in your clusters | Through KubeOn: GPU utilization, idle GPU cost and node sizing |
Capacity planning without the spreadsheet
Per-minute sizing
Percentiles from per-minute utilization over 30 days.
Pool-aware coverage
Global, Data Zone and Regional kept apart.
Break-even analysis
When PTU beats pay-as-you-go, and for how long.
Renewal timeline
Expiring reservations matched to current needs.
Approval-gated purchases
Nothing is bought without a named approver.
Past billing corrected
Hourly PTU separated from pay-as-you-go history.
Provisioned capacity: your questions answered
UnitVane AI
Keep exploring
Evaluate provisioned capacity with your team
Bring your platform and FinOps priorities. We will walk through the data, controls and evidence for your evaluation.