ACS charges for two types of resources: container computing power and the cloud services your clusters depend on. There is no cluster management fee.
ACS container computing power fees
Billing is per pod instance, pay-as-you-go. Charges are calculated per second and invoiced hourly.
Billing formula
ACS pod instance:
Fee = (Number of vCPUs × vCPU unit price + Memory size × Memory unit price) × Billing duration
GPU pod:
Fee = (Number of vCPUs × vCPU unit price + Memory size × Memory unit price + Number of GPUs × GPU unit price) × Billing duration
Billing method
You are billed per pod instance on a pay-as-you-go basis. Billing is calculated per second and invoiced hourly.
General-purpose computing power billing
Region | Instance type | Compute quality | Architecture | Billing items (per second) | |
vCPU | Memory (GiB) | ||||
Chinese mainland | General-purpose | default | Intel | USD 0.0000026/second (USD 0.00936/hour) | USD 0.00000128/second (USD 0.0046/hour) |
best-effort | USD 0.00000052/second (USD 0.001872/hour) | USD 0.00000025/second (USD 0.0009/hour) | |||
Compute-optimized | default | USD 0.0000034/second (USD 0.01224/hour) | USD 0.00000167/second (USD 0.006/hour) | ||
best-effort | USD 0.00000068/second (USD 0.002448/hour) | USD 0.000000334/second (USD 0.0012024/hour) | |||
Hong Kong (China) and outside China | General-purpose | default | USD 0.00000381/second (USD 0.013716/hour) | USD 0.00000192/second (USD 0.0069/hour) | |
best-effort | USD 0.00000076/second (USD 0.002736/hour) | USD 0.00000038/second (USD 0.001368/hour) | |||
Compute-optimized | default | USD 0.00000496/second (USD 0.017856/hour) | USD 0.00000248/second (USD 0.008928/hour) | ||
best-effort | USD 0.000000992/second (USD 0.0035712/hour) | USD 0.000000496/second (USD 0.0017856/hour) | |||
For example, a general-purpose instance with default compute quality in the Chinese mainland has 2 vCPUs and 8 GiB of memory. Billing starts at 10:05, and an invoice is generated at 11:00. The fee is (2 vCPUs × USD 0.0000026/second + 8 GiB × USD 0.00000128/second) × 3300 seconds, which is approximately USD 0.05.
Accelerated computing power billing
Region | Compute type | Compute quality | Billing items (per second) | |||
vCPU | Memory (GiB) | GPU (card) | ||||
Chinese mainland | gpu | default | USD 0.0000026/second (USD 0.00936/hour) | USD 0.00000128/second (USD 0.004608/hour) | T4 (16 GiB GPU memory) | USD 0.00025889/second (USD 0.932/hour) |
A10 (24 GiB GPU memory) | USD 0.00028833/second (USD 1.038/hour) | |||||
best-effort | USD 0.00000104/second (USD 0.003744/hour) | USD 0.00000051/second (USD 0.0018432/hour) | T4 (16 GiB GPU memory) | USD 0.00010356/second (USD 0.3728/hour) | ||
A10 (24 GiB GPU memory) | USD 0.00011533/second (USD 0.4152/hour) | |||||
Hong Kong (China) and outside China | gpu | default | USD 0.00000381/second (USD 0.0137/hour) | USD 0.00000192/second (USD 0.0069/hour) | T4 (16 GiB GPU memory) | USD 0.00012222/second (USD 0.44/hour) |
A10 (24 GiB GPU memory) | USD 0.00039722/second (USD 1.43/hour) | |||||
best-effort | USD 0.00000152/second (USD 0.0055/hour) | USD 0.00000077/second (USD 0.0028/hour) | T4 (16 GiB GPU memory) | USD 0.00004889/second (USD 0.176/hour) | ||
A10 (24 GiB GPU memory) | USD 0.00015889/second (USD 0.572/hour) | |||||
For example, a GPU instance with default compute quality in the Chinese mainland has 8 vCPUs, 32 GiB of memory, and 1 T4 GPU. Billing starts at 10:05, and an invoice is generated at 11:00. The fee is (8 vCPUs × USD 0.0000026/second + 32 GiB × USD 0.00000128/second + 1 GPU × USD 0.00025889/second) × 3300 seconds, which is approximately USD 1.06.
Billing duration
Billing starts when the container image begins downloading (Pending state) and stops when the instance finishes running (Succeeded or Failed state). Instances that fail to create or start are not charged.
General-purpose computing power
Prices are per second. The hourly equivalent is shown in parentheses.
Chinese mainland
| Instance type | Compute quality | Architecture | vCPU | Memory (GiB) |
|---|---|---|---|---|
| General-purpose | default | Intel | USD 0.0000026/s (USD 0.00936/hr) | USD 0.00000128/s (USD 0.0046/hr) |
| General-purpose | best-effort | Intel | USD 0.00000052/s (USD 0.001872/hr) | USD 0.00000025/s (USD 0.0009/hr) |
| Compute-optimized | default | Intel | USD 0.0000034/s (USD 0.01224/hr) | USD 0.00000167/s (USD 0.006/hr) |
| Compute-optimized | best-effort | Intel | USD 0.00000068/s (USD 0.002448/hr) | USD 0.000000334/s (USD 0.0012024/hr) |
Hong Kong (China) and outside China
| Instance type | Compute quality | vCPU | Memory (GiB) |
|---|---|---|---|
| General-purpose | default | USD 0.00000381/s (USD 0.013716/hr) | USD 0.00000192/s (USD 0.0069/hr) |
| General-purpose | best-effort | USD 0.00000076/s (USD 0.002736/hr) | USD 0.00000038/s (USD 0.001368/hr) |
| Compute-optimized | default | USD 0.00000496/s (USD 0.017856/hr) | USD 0.00000248/s (USD 0.008928/hr) |
| Compute-optimized | best-effort | USD 0.000000992/s (USD 0.0035712/hr) | USD 0.000000496/s (USD 0.0017856/hr) |
Example: A General-purpose default instance in the Chinese mainland with 2 vCPUs and 8 GiB of memory, running from 10:05 to the 11:00 invoice cutoff (3,300 seconds):
(2 × USD 0.0000026 + 8 × USD 0.00000128) × 3,300 ≈ USD 0.05
Accelerated computing power (GPU)
Chinese mainland
| Compute type | Compute quality | vCPU | Memory (GiB) | GPU card |
|---|---|---|---|---|
| gpu | default | USD 0.0000026/s (USD 0.00936/hr) | USD 0.00000128/s (USD 0.004608/hr) | T4 (16 GiB GPU memory): USD 0.00025889/s (USD 0.932/hr) |
| gpu | default | — | — | A10 (24 GiB GPU memory): USD 0.00028833/s (USD 1.038/hr) |
| gpu | best-effort | USD 0.00000104/s (USD 0.003744/hr) | USD 0.00000051/s (USD 0.0018432/hr) | T4 (16 GiB GPU memory): USD 0.00010356/s (USD 0.3728/hr) |
| gpu | best-effort | — | — | A10 (24 GiB GPU memory): USD 0.00011533/s (USD 0.4152/hr) |
Hong Kong (China) and outside China
| Compute type | Compute quality | vCPU | Memory (GiB) | GPU card |
|---|---|---|---|---|
| gpu | default | USD 0.00000381/s (USD 0.0137/hr) | USD 0.00000192/s (USD 0.0069/hr) | T4 (16 GiB GPU memory): USD 0.00012222/s (USD 0.44/hr) |
| gpu | default | — | — | A10 (24 GiB GPU memory): USD 0.00039722/s (USD 1.43/hr) |
| gpu | best-effort | USD 0.00000152/s (USD 0.0055/hr) | USD 0.00000077/s (USD 0.0028/hr) | T4 (16 GiB GPU memory): USD 0.00004889/s (USD 0.176/hr) |
| gpu | best-effort | — | — | A10 (24 GiB GPU memory): USD 0.00015889/s (USD 0.572/hr) |
Example: A GPU default instance in the Chinese mainland with 8 vCPUs, 32 GiB of memory, and 1 T4 GPU, running from 10:05 to the 11:00 invoice cutoff (3,300 seconds):
(8 × USD 0.0000026 + 32 × USD 0.00000128 + 1 × USD 0.00025889) × 3,300 ≈ USD 1.06
Configuration adjustments
When the vCPU or memory you specify is not in a supported configuration, ACS adjusts it to the nearest supported configuration and bills for the adjusted values. For the list of supported configurations, see Billing methods.
Cloud service resource fees
The following Alibaba Cloud services can be used with ACS. Each service is billed independently according to its own pricing rules.
| Service | Activation type | Description | Subscription | Resource plans | Billing details |
|---|---|---|---|---|---|
| Virtual Private Cloud (VPC) | Required | Builds the cluster network environment and routing rules. For more information, see What is a virtual private cloud (VPC)?. | Not supported | Not supported | Product billing |
| Server Load Balancer | Recommended | Creates load balancers for your cluster. For more information, see What is Classic Load Balancer (CLB)? and What is Application Load Balancer (ALB)?. | CLB: Not supported; ALB: Not supported | CLB: Supported; ALB: Supported | CLB: Pay-as-you-go; ALB: ALB billing rules |
| Container Registry (ACR) | Recommended | Provides secure hosting and full lifecycle management for cloud-native assets. For more information, see What is Container Registry (ACR)?. | Supported | Not supported | Billing overview |
| Simple Log Service (SLS) | Recommended | Collects and retrieves logs from cluster components and applications. For more information, see What is Simple Log Service?. | Not supported | Supported | Billing overview; Purchase a resource plan |
| Elastic IP Address (EIP) | Optional | Enables communication between cloud resources and the Internet. For more information, see What is an Elastic IP Address (EIP)?. | Not supported | Not supported | Billing overview |
| NAT Gateway | Optional | Enables public network access and public network image pulls for your cluster. For more information, see What is NAT Gateway?. | Not supported | Supported | NAT Gateway billing; NAT Gateway resource plans |
| File Storage NAS | Optional | Provides file storage for cluster application data using NAS. For more information, see What is File Storage NAS?. | Not supported | Supported | General-purpose NAS billing; Extreme NAS billing; Purchase a resource plan |
| Object Storage Service (OSS) | Optional | Provides object storage for cluster application data. For more information, see What is Object Storage Service (OSS)?. | Not supported | Supported | Billing overview |
| Key Management Service (KMS) | Optional | Manages keys for cluster applications and enables encryption at rest for clusters. For more information, see What is Key Management Service (KMS)?. | Not supported | Not supported | KMS 1.0 billing overview |
| Alibaba Cloud DNS PrivateZone | Optional | Provides private DNS service for your VPC environment. For more information, see What is PrivateZone?. | Not supported | Not supported | Pricing |
ACR enterprise instances support only subscription billing. For more information, see Billing for ACR enterprise instances.
FAQ
When does billing start?
Billing starts when ACS begins pulling the container image for your pod instance (Pending state). Charges begin before the container is actually running.
When does billing stop?
Billing stops when the instance reaches the Succeeded or Failed state. No further charges accrue after the instance stops running.
What if my pod fails to start?
If an instance fails to create or start, no charges apply.
What happens if I specify an unsupported vCPU or memory configuration?
ACS automatically adjusts the configuration to the nearest supported option and bills for the adjusted values, not your original request. Check the supported configurations in Billing methods before creating instances to avoid unexpected costs.