AI Gateway Serverless is billed based on instance type and actual usage. Fees consist of the instance fee and pay-as-you-go billable items. For specific purchasable instance types, prices, discounts, billing cycles, and available regions, refer to the product console, official pricing page, and the Expenses and Costs page.
Scope
This topic describes the commercial billing rules of AI Gateway Serverless on the international site. It applies to AI Gateway Serverless (Serverless Standard Edition and Serverless Enterprise Edition), and does not apply to Serverless (legacy) instances or dedicated instance editions.
Fees for other cloud products, such as model services, Simple Log Service (SLS), and CloudMonitor (CMS), are billed separately by the corresponding cloud products.
Billing differences by instance type
AI Gateway Serverless includes Serverless Standard Edition and Serverless Enterprise Edition. The two instance types use different billing methods:
Comparison item | Serverless Standard Edition | Serverless Enterprise Edition |
Starting cost | Starts at USD 0, with no instance fee | Instance fee applies |
Billing method | Pay-as-you-go billing method based on actual usage | Instance fee plus pay-as-you-go billing method based on actual usage |
Entry capability | Uses a platform-assigned entry point | Provides independent entry capability |
Custom domain name and encryption in transit configuration | Does not support custom user configuration | Supports custom domain names, certificates, TLS policy, and HTTP/2 configuration |
Internet traffic | Outbound internet traffic and inbound internet traffic are billed by the gateway at USD 0.118/GB | Outbound internet traffic and inbound internet traffic are billed by the gateway at USD 0.118/GB |
Default QPS limit | 50 QPS by default | 500 QPS |
Service level | 99.9% | 99.95% |
If you want to try the service at low cost, run development tests, or support low-traffic workloads, you can choose Serverless Standard Edition. If your workloads are in production, or if you require an independent entry point, custom domain names, and a higher service level, we recommend Serverless Enterprise Edition.
Billing items
AI Gateway Serverless is billed based on instances and actual usage. The following billable items apply:
Billing item | Billing dimension | Billing method | Reference unit price |
Instance fee | Instance runtime | Charged based on instance runtime for Serverless Enterprise Edition only. Serverless Standard Edition has no instance fee. | USD 0.0365/hour |
Request calls | Number of requests, in units of 10,000 calls | Tiered pricing based on request calls that pass through the gateway. Higher usage has a lower unit price. The tiers are described in the following table. Persistent connections are converted into request calls based on connection duration, as described later in this topic. | USD 0.0012 to USD 0.0022 per 10,000 calls |
Data processing volume | Data volume in GB | Billed hourly based on the sum of request body and response body data volumes, regardless of whether the traffic uses the Internet or a private network. | USD 0.005/GB/hour |
Consumer rule expansion count | Number of rules | Charged based on the purchased expansion quantity when the consumer-related quotas exceed their default quotas. After you purchase consumer rule expansion capacity, the expansion quantity raises the upper limits of the following three consumer-related quotas at the same time: Consumer authorization rules, Number of consumers, and Consumer quota rules. For the default quotas, see Instance quotas. The maximum quota increase available for purchase is subject to the purchase page. | USD 0.003 per 100 rules/hour |
Outbound internet traffic | Outbound public network traffic | Billed by the gateway. If you configure your own NAT Gateway and outbound internet traffic goes through your NAT Gateway, the gateway does not charge this traffic. | USD 0.118/GB |
Inbound internet traffic | Inbound public network traffic | Billed by the gateway. | USD 0.118/GB |
Request calls, data processing volume, and consumer rule expansion count use the same billing methods and unit prices for Serverless Standard Edition and Serverless Enterprise Edition. Final fees may vary based on regions and account-level discounts. The preceding unit prices are for reference only. Actual unit prices are subject to the official pricing page and your bills.
Internet traffic
Outbound internet traffic and inbound internet traffic are billed by the gateway at USD 0.118/GB:
Outbound internet traffic: traffic generated when the gateway accesses public services. If you configure your own NAT Gateway and outbound internet traffic goes through your NAT Gateway, the gateway does not charge this traffic.
Inbound internet traffic: traffic generated when the public-facing endpoint of the gateway is accessed.
Private network access does not incur Internet traffic fees.
Tiered pricing for request calls
Request calls are priced by hourly usage tiers. Higher usage has a lower unit price:
Hourly call volume | Reference unit price |
0 to 500,000 calls | USD 0.0022 per 10,000 calls |
500,000 to 1,000,000 calls | USD 0.0018 per 10,000 calls |
More than 1,000,000 calls | USD 0.0012 per 10,000 calls |
Measurement rules for request calls
Request calls are measured by the number of requests that pass through the gateway. Persistent connections, such as streaming responses, MCP persistent connections, and Agent persistent sessions, continuously occupy gateway resources while the connection is maintained. Therefore, persistent connections are converted into request calls based on connection duration:
Measurement rule | Description |
Short-lived connection request | A standard request-response request is counted as one request call. |
Persistent connection measured by duration | A persistent connection is counted as one request call for every 30 seconds of connection duration. |
Measurement examples:
Request characteristic | Measured request calls |
Standard AI conversation request that completes within seconds | 1 |
Persistent connection maintained for 2 minutes | 4 |
Persistent connection maintained for 10 minutes | 20 |
Request body and response body data volumes are billed separately as data processing volume, and do not affect the number of request calls. Specific measurement rules and unit prices are subject to the official pricing page, product console, and your bills.
Fee calculation method
Gateway fee = Instance fee + Request call fee + Data processing volume fee + Rule expansion fee + Internet traffic feeOutbound internet traffic and inbound internet traffic are billed in the gateway bill at USD 0.118/GB.
AI-enhanced capabilities such as intelligent routing and model token fees are billed separately.
View bills
You can view the usage and billing details of AI Gateway Serverless on the Expenses and Costs page.