Billing
Overview of Billing Components
Serverless Kubernetes is billed per-minute and based on the resources your pods use. Here are the key components of our billing model:
- GPU: Each GPU used in your pods is billed at a fixed rate.
- vCPU: Every vCPU used by your pods is billed. If your pods use CPU resources without a GPU, these are billed separately. CPU usage is billed at 1/100th the cost of a vCPU.
- Memory: Memory usage is billed per MB. This allows for granular control over the memory allocation and cost.
- Load Balancer: Running Load balancer on the cluster.
Billing for Suspended/Stopped/Idle Resources
Pods are only billed when they are actively running. If a pod is stopped, suspended, or idle, you will not be charged for those resources during that time.
When a cluster is Suspended, all pods are stopped and stop being billed for. The exception to this are load balancers, as the IP address is persisted during the time the cluster is suspended.
Resource Pricing
The following tables outlines the pricing per hour/per minute for:
1. GPU Resources:
| Resource | SKU | Price ($/minute) | Price ($/hour) |
|---|---|---|---|
| NVIDIA H100 SXM 80GB | ogx-k8s-h100sxm-80 | 0.0855 | 5.13 |
| NVIDIA H100 80GB | ogx-k8s-h100 | 0.069408 | 4.16448 |
| NVIDIA L4 24GB | ogx-k8s-l4 | 0.019044 | 1.14261 |
| NVIDIA L40S 48GB | ogx-k8s-l40s | 0.036000 | 2.16000 |
2. Other Resources:
| Resource | SKU | Price ($/minute) | Price ($/hour) |
|---|---|---|---|
| Memory (MB) | ogx-k8s-mem | 0.00000008 | 0.000005 |
| CPU (1/100) | ogx-k8s-vcpu | 0.00000167 | 0.0001 |
| Load Balancer | ogx-k8s-lb | 0.0003015 | 0.01809 |
You can read more about billing in our Support section.