Estimate your bill
CPU
Capacity AI allocates a continuously optimized number of vCPU millicores (1/1000 of a vCPU) and RAM to guarantee a significant cost saving.
RAM
Capacity AI reduces RAM consumption by approx. 10x, achieving major savings.
Volumes
Control Plane offers optional persistent storage for your workloads. You pay per GB-month of provisioned storage.
Additional CPU and memory charges apply per shared volume mount point.
Observability
Pay per GB-month of storage. The first 100 GB-months are free.
First 50M samples free, then $0.5 per 10M samples.
Egress traffic
Egress traffic represents the amount of data that is sent to the internet from your workloads.
Dedicated load balancing
For advanced use cases, you can provision a dedicated load balancer per GVC™. You pay per load balancer-month.
Pay-as-you-go GPU compute
| GPU | Memory | Per GPU-hour |
|---|---|---|
| NVIDIA B200 | — | $7.93 |
| NVIDIA H200 | 141 GB | $4.44 |
| NVIDIA H100 | 80 GB | $4.18 |
| NVIDIA A100 | 80 GB | $2.23 |
| NVIDIA A100 | 40 GB | $2.02 |
| NVIDIA V100 | 16 GB | $1.59 |
| NVIDIA A10 | 24 GB | $1.29 |
| NVIDIA A6000 | 48 GB | $1.15 |
| NVIDIA A40 | 48 GB | $0.82 |
| NVIDIA L40S | 48 GB | $0.79 |
| NVIDIA A5000 | 24 GB | $0.77 |
| NVIDIA A4000 | 16 GB | $0.69 |
| NVIDIA L4 | 24 GB | $0.44 |
| NVIDIA T4 | 16 GB | $0.35 |
What you don't pay for
- NAT gateways
- Internet gateways
- TLS certificates
- DNS and global DNS
- Secrets storage and management
- Container registry
- Log storage and consolidation
- Metrics (first 50M samples)
- Shared load balancing
- Audit trail
- Container orchestration (Kubernetes)
- Service mesh (Istio)
- VPN between VPCs
- Auto-scaling
- Geo-routing & failover
Pricing comparison calculator
Already running a Kubernetes cluster? Use our open-source K8s cost analyzer to calculate what the same workloads would cost on Control Plane versus your cloud provider.
Get the K8s cost analyzer
