Billing FAQ

Ask PAI Agent first

The PAI AI assistant (Xiao PAI) covers the full PAI product suite — providing usage guidance, and offering diagnostic capabilities for DSW instances, DLC jobs, and EAS services. It automatically identifies failure causes and recommends next steps and tools.

Q: Which PAI services incur charges?

Running DSW (Data Science Workshop) instances consume computing resources and incur charges. Stop unused instances to avoid unnecessary costs.

Important
After stopping an instance, you are still charged for its system disk if you have scaled it out. To stop all billing for a DSW instance, you must delete it. Before you delete the instance, back up all necessary data.
DLC tasks that run in a public resource group incur charges. Tasks that run in a dedicated resource group do not incur additional charges because these groups are subscription-based.
For EAS (Elastic Algorithm Service) services in a public resource group, billing starts when the instance provisions resources, even if the service receives no calls. For services in a pay-as-you-go dedicated resource group, the resource group itself incurs charges. For services in a subscription dedicated resource group, no additional charges apply.

The following are basic billing rules. For more information, see EAS billing.
- A pay-as-you-go dedicated resource group incurs charges when it is in the Running state.
  
  Important
  During scaling (out or in), the resource group enters an intermediate state, such as Scaling Out or Scaling In. You are still charged for provisioned resources during this time.
- For services deployed in a public resource group, billing starts as soon as the instance provisions resources.
  
  Important
  When a deployed service scales out, it enters an intermediate state (Waiting). You are still charged for the provisioned resources during this time.
- System disk charges.
  - If you add a system disk when you deploy an EAS service in a public resource group, you are still charged for the system disk after the service is stopped. To stop system disk charges, delete the EAS service. Before you delete the service, make sure it is no longer in use.
  - If you add a system disk when you purchase a machine for a dedicated resource group, billing for the system disk stops only after the pay-as-you-go machine is deleted or the subscription expires. Before you delete the machine, make sure it is no longer in use.
Running workflows in Machine Learning Designer / Studio incur charges for the algorithm components used during computation.
To check your free trial usage, validity period, and other details, go to the My Trials page in the Expenses and Costs console.

Q: How to stop a billable item?

Stop a DSW instance.
Important
- If a DSW instance is created by using a dedicated resource group and no datasets are mounted to the instance, the system disk of the DSW instance is used for temporary storage. If you stop or delete the DSW instance, data cannot be restored. Proceed with caution.
- If a DSW instance is created by using a public resource group, the environment is saved on the free disk mounted to the instance after you stop the instance. If you do not restart the instance within 15 days after the instance is suspended, the disk is released.
- If you resize a system disk when you create a DSW instance, you are still charged for the system disk even if the instance is stopped. If you want to stop all billing related to the DSW instance, delete the DSW instance. Make sure that the required data is backed up before you delete the instance.
- Create DSW instances within a workspace to help prevent accidental operations that can lead to unnecessary charges.
1. Log on to the PAI console. In the top navigation bar, select the target region and, on the right, select the target workspace. Then, click Enter DSW.
2. On the Interactive Modeling (DSW) page, find the target instance and click Stop in the Actions column.
  
  After the instance is stopped, its Status changes to Stopped. At this point, billing stops for pay-as-you-go instances. When you exit DSW, make sure your instance is in the Stop state to avoid unexpected charges.
Stop billing for an EAS service:

Important
To avoid business disruptions, make sure the services you stop are no longer needed.
- Stop services running in a pay-as-you-go dedicated resource group.
  
  To stop billing, reduce the server count in the dedicated resource group to 0:
  1. Log on to the PAI console. Select a region on the top of the page. Then, select the desired workspace and click Elastic Algorithm Service (EAS).
  2. On the Elastic Algorithm Service (EAS) page, click the Resource Group tab. Click the name of the target resource group to go to its details page.
  3. On the Machine List tab, find the pay-as-you-go machine and click Delete in the Actions column. This deletes the machine and stops billing.
- Stop services running in a public resource group:
  
  On the Elastic Algorithm Service (EAS) page, click the Inference Service tab. Find the target service and click Stop in the Actions column. This stops the service and its billing.
  
  Note
  If you added a system disk when you deployed the EAS service, you are still charged for the system disk even after the service is stopped. To stop all billing related to the EAS service, you must delete the service. Before deleting the service, ensure it is no longer needed to prevent business disruptions.
- Stop pay-as-you-go billing for an EAS dedicated gateway
  
  On the Elastic Algorithm Service (EAS) page, click the Dedicated Gateway tab. Find the pay-as-you-go EAS dedicated gateway and click Delete in the Actions column.
Unsubscribe from an AI Computing Resource Group (Subscription), EAS dedicated gateway (Subscription), or EAS dedicated machine (Subscription)

Important
To avoid business disruptions, make sure the resources you unsubscribe from are no longer needed.
1. Log on to the Expenses and Costs console.
2. In the navigation pane on the left, choose Order > Unsubscription.
3. Select an unsubscription type. For Product Name, select China Site_AI Computing Resource Group (Subscription), EAS dedicated gateway (Subscription), or EAS dedicated machine (Subscription), and follow the on-screen instructions to complete the process.
Stop a Machine Learning Designer / Studio billable item.
1. Log on to the PAI console. In the top navigation bar, select the target region and, on the right, select the target workspace. Then, click Go to Designer.
2. On the Visualized Modeling (Designer) page, double-click the workflow name to open the workflow page.
3. On the canvas, click the button to stop the running workflow.

Q: Locating DSW, DLC, or EAS services

On the overview page, verify the region of your DSW, DLC, or EAS resources and switch to the correct one if needed.

Q: How to view charges and billing details?

On the Billing Details page, filter bills to view detailed billing information for machine learning. For more information, see View billing details. The Product Details column indicates the source of the charge. Possible values include:

EAS Dedicated Machine Pay-as-you-go: Charges incurred by a pay-as-you-go EAS dedicated resource group.

Machine Learning Platform for AI (PAI): Charges for using MaxCompute computing resources for training in Machine Learning Designer / Studio. Use the Billable Item column to identify specific charges. The following table describes the mappings.

Product detail	Billable item	Instance ID	Source of charge
Machine Learning (PAI)	Usage	text_analysis data_analysis data_manipulation deep_learning default	Charges generated by experiment training in Machine Learning Designer / Studio.

EAS Dedicated Machine (Subscription): Charges incurred by a subscription EAS dedicated resource group.

Q: Still charged after stopping a service?

Billing delay: For pay-as-you-go DSW or EAS resources, bill generation and payment deduction are delayed. The charges you see may be for usage before the resource was stopped, not after.
DSW instance system disk: If you scaled out the system disk when you created the instance, you are still charged for the disk after the instance is stopped. To stop all charges, delete the instance. Before deleting the instance, back up all necessary data.
EAS service running in the background: An unused service is not a stopped service. It continues to consume resources and incur charges until you stop or delete it. Before deleting the service, back up all necessary data.

Q: Extra charges on EAS trial resources?

Reason for charges: When you deploy stable-diffusion-webui or comfyui in EAS, if the selected image version has a -api or -cluster suffix, the deployment uses an asynchronous queue resource. This resource is billable and is not covered by EAS trial resources.
Solution: When you deploy stable-diffusion-webui or comfyui, select a standard image version, which either has the -standard suffix or no suffix.

Q: Still charged after stopping an item?

Alibaba Cloud billing is not processed in real time. A delay exists between resource usage and bill generation. For example, charges for resources used between 10:00 and 11:00 might be billed several hours later. Even if you stop a resource at 11:00, you might still receive a deduction notice hours later. This delay can create the impression of being overcharged, but it is standard billing behavior, not an actual overcharge.

Q: Extra charges with a DSW resource plan?

Reason for charges: If you use a DSW resource plan but still incur extra charges, your DSW instance likely continued running after the plan expired. The instance then generates pay-as-you-go charges. Due to billing delays, these charges may appear later.
Solution:
- You can go to the Manage Instance page in the Expenses and Costs console to check the expiration date of your DSW resource plan and purchase a new one before it expires.
- Maintain a sufficient balance in your Alibaba Cloud account to avoid service disruptions from overdue payments after your resource plan expires.

Q: Charges for Designer with no running workflows?

Machine Learning Designer charges are not limited to workflows created within the platform. Charges can also come from the following sources:

odpscmd tool: You can run PAI commands to execute specific algorithms by accessing MaxCompute with the odpscmd tool.
DataWorks: You can run PAI commands in an ODPS SQL node to execute specific algorithms. For an example, see Video-based Personalized Recommendation (Collaborative Filtering)-rec_test_etrec_i2i_output_20230216_v1.

Check these sources to identify the cause of the charges.

Q: Why are bills paid by stored-value card?

Alibaba Cloud deducts pay-as-you-go bills from payment sources in the following order:

Valid coupons in your Alibaba Cloud account.
Valid stored-value cards in your Alibaba Cloud account.
The cash balance in your Alibaba Cloud account. Funds transferred from other accounts are used before your own top-ups.

If your account has a valid and applicable coupon, charges are deducted from the coupon first. If no applicable coupon exists, charges are deducted from a stored-value card. If the stored-value card balance is insufficient, charges are deducted from your cash balance. For more information, see Deduction order for pay-as-you-go bills.

Q: How to check plan usage?

You can log on to the Expenses and Costs console to view the usage of your savings plans or resource plans.

Q: What should I do if the "Failed to call the billing system" error appears when I activate the PAI labeling platform or a workspace?

This error is usually caused by an exception in the interaction with the billing system. It does not mean that the activation itself failed.

Activating a workspace is free of charge. Charges are incurred only when you use specific modules such as DSW, DLC, and EAS, and each module is billed according to its own billing rules.

Check the status of your Alibaba Cloud account first, for example, whether payments are overdue. After you resolve the issue, activate the service again.

Q: Are pay-as-you-go DSW charges refundable? Can I settle the bill manually in advance?

A pay-as-you-go DSW instance starts to incur charges as soon as its status changes to Running. Charges are incurred even if you perform no operations, and these charges are not refundable. To stop billing, delete the instance. Before you delete the instance, back up all necessary data.

Pay-as-you-go bills are automatically deducted from your Alibaba Cloud account balance at the beginning of the following month. Manual settlement in advance is not supported.

Q: Why does the DSW instance price returned by OpenAPI differ from the price shown in the console?

The price difference may be caused by the following reasons:

Currency difference: The console may display prices in US dollars (USD), whereas OpenAPI returns prices in the local currency.
Mismatched query parameters: The region, billing method, or instance type in the query does not match the options selected in the console.
Different pricing scope: OpenAPI returns the list price, which is not a real-time price, whereas the console may display a bundled package price or a discounted price.

For actual purchases and billing, the price shown in the console prevails.

Q: How do I confirm whether my account has activated PAI or has trial resources? Where can I view the quota?

Where to check: Go to the My Trials page in the Expenses and Costs console. This page shows the status, validity period, and quota usage of all trial resource plans under your account in one place.
How quotas are displayed: After you activate the DSW free trial, the resource quota is not displayed directly in the instance list of the PAI console. A deduction is triggered only when you create an instance, select a specification that is marked as a free trial, and your account still has valid free trial eligibility. For routine queries, use the data in the Expenses and Costs console.
Confirming activation status: If the My Trials page shows only trial resource plans and no other PAI product orders, you have claimed only the trial benefits and have not activated a paid service. No additional deactivation is required.

Q: Can auto-renewal be canceled for the trial resource plan used in training scenarios for university teachers and students?

The trial resource plan dedicated to training scenarios for university teachers and students does not support auto-renewal, and no auto-renewal feature exists for it. The resource plan expires automatically at the end of its validity period. You do not need to cancel renewal or release the plan manually.

Q: What is the price difference between a PAI-EAS GPU instance and an ECS GPU instance of the same model?

The price of PAI-EAS GU30-series GPU instance types is more than 20% lower than that of traditional A10 instance types.
Except for the GU30 series, no fixed price difference ratio applies to other instance types. For the exact price, refer to the specification price displayed when you make a purchase in the console of each product.
To view ECS GPU specifications and prices, go to the ECS console or the ECS pricing page. The PAI console displays only the specifications available to EAS and cannot display ECS GPU information.

Q: Why is my EAS pay-as-you-go bill not offset even though my PAI savings plan still has a quota?

The possible reasons are as follows:

Bill delay: The generation and deduction of EAS pay-as-you-go bills are delayed. The bill that is currently not offset may cover charges that were incurred before the savings plan took effect or before the instance was released. Wait for subsequent bills to be updated and then check again.
Resource type mismatch: A savings plan offsets only the pay-as-you-go charges of an EAS public resource group. Charges for EAS dedicated resources, general computing resources 2.0, or other non-public resources cannot be offset.

To check this yourself, log on to the Expenses and Costs console and view the remaining quota and validity period of the savings plan on the My Subscriptions page. In the bill details, filter by the product name EAS Service Pay-as-you-go to verify the resource type.

Q: Can data-dependent components such as databases be deployed in a PAI-PPU environment?

PAI-PPU supports the deployment of data-dependent components such as databases through Kubernetes or PAI products. If model training or inference depends on data, configure storage and data loading as described in the PAI-PPU best practices documentation.

The deployment compatibility of a specific database component must be evaluated and confirmed based on your business scenario, such as I/O performance requirements and network topology.