Amazon API Gateway pricing depends on the API type and the traffic it handles. HTTP and REST APIs charge for API request units and outbound data; HTTP request payloads are metered in 512 KB increments. WebSocket APIs charge for messages and connection minutes. REST API caching and private connectivity can add separate charges, and the backend services you invoke have their own prices.

The rates and AWS examples below were checked against the Amazon API Gateway pricing page on October 3, 2026. AWS prices can vary by Region. Treat each example as an estimate for its stated assumptions, then check the current price for the Region where your API runs.

API Gateway costs at a glance

API type What API Gateway meters Published rate example and Region
HTTP API Incoming request units, metered in 512 KB payload increments, plus data transferred out US East (N. Virginia, us-east-1): $1.00 per million units for the first 300 million; $0.90 per million above that tier
REST API API calls and data transferred out; optional stage cache charged by cache size and time US East (N. Virginia and Ohio), US West (Oregon), and Asia Pacific (Mumbai): $3.50 per million for the first 333 million calls; $2.80 per million for the next 667 million
WebSocket API Messages sent and received, plus connected minutes In US East (N. Virginia), $1.00 per million messages and $0.25 per million connection minutes

These are usage charges for API Gateway. Outbound data, a REST cache, AWS PrivateLink, Lambda, logging, and other services can change the total. See AWS's current rate table and Region-specific examples before budgeting production traffic.

Request and data charges by API type

HTTP APIs

HTTP API request payloads are metered in 512 KB increments. A request with a payload up to 512 KB counts as one request unit; a larger payload can count as several units. For example, AWS prices a 4.5 MB upload as nine 512 KB request units. AWS's current pricing page shows this metering in its upload example; its API Gateway private APIs and private integration whitepaper identifies the HTTP request tiers for us-east-1. The number of HTTP calls alone is therefore not enough to estimate a workload that sends large request bodies.

AWS currently lists $1.00 per million HTTP request units for the first 300 million in a month and $0.90 per million above that tier in US East (N. Virginia). Outbound data is a separate cost. Small responses do not remove the request charge, and large responses can make transfer charges material. Check the live price for another Region.

Example: Suppose an API receives 10 million requests in a month and each request payload is no larger than 512 KB. At the first listed tier, the API Gateway request charge is 10 × $1.00 = $10. This is the request portion only; add applicable data transfer and integration costs. A request with a 1 MB payload counts as two request units, so 10 million such requests would produce 20 million billable units.

REST APIs

Standard REST API usage is charged per API call, with separate charges for outbound data. AWS's published examples show the first request tier at $3.50 per million calls for the first 333 million per month, followed by $2.80 per million for the next 667 million in US East (N. Virginia and Ohio), US West (Oregon), and Asia Pacific (Mumbai). At higher volumes, AWS lists additional lower tiers; use the live pricing page for the complete schedule and your Region.

AWS's published example: Five million Regional or edge-optimized REST API calls in US East (N. Virginia or Ohio), US West (Oregon), or Asia Pacific (Mumbai), each returning a 3 KB response with caching off, produce $17.50 in call charges. AWS calculates 14.3 GB of outbound data at $0.09 per GB, or $1.29. The published API Gateway total is $18.79. The example does not include the backend, logs, or other services.

For REST APIs with response streaming enabled, large response bodies affect the billable-request count. AWS divides the response payload by 10 MB and rounds up, with at least one billable request unit per API call. For example, its 65 MB response example counts seven units. The response data transfer charge still applies, so include both parts when estimating large downloads.

REST API cache charges

API Gateway's optional stage cache is available for REST APIs and is charged hourly by cache size. AWS lists a 1.6 GB cache at $0.038 per hour in its example for US East, US West, and EU (Ireland). If it remains on for a 730-hour month, that is about $27.74 before other charges. Check the price for your Region and cache size.

A cache can reduce backend requests and latency, but API Gateway still charges for each API call that reaches it, including cache hits, and the response is still transferred to the client. Caching therefore does not remove API Gateway request or outbound-data charges. Compare the cache's hourly cost only with backend work it can avoid. Keep cache keys, time-to-live, and invalidation behavior correct for the data being served. See AWS's REST API cache guidance.

WebSocket APIs

WebSocket APIs charge for each message sent or received and for the total connected minutes. Messages are metered in 32 KB increments: a 33 KB message counts as two billable messages. A message can be up to 128 KB. WebSocket control frames such as ping and pong are not metered. These message units differ from HTTP or REST API calls.

In US East (N. Virginia), AWS's published rates are $1.00 per million messages and $0.25 per million connection minutes.

AWS's published chat example: 1,000 users each stay connected for 12 hours a day for 30 days. Each user sends 100 messages and receives 500 messages a day, all 3 KB or smaller:

  • Messages: (100 + 500) × 1,000 × 30 = 18 million; 18 × $1.00 = $18.00.
  • Connection time: 1,000 × 12 × 60 × 30 = 21.6 million minutes; 21.6 × $0.25 = $5.40.
  • API Gateway total: $18.00 + $5.40 = $23.40, before integrations and other services.

Private APIs and PrivateLink

API Gateway does not charge outbound data transfer for a Private API, but that does not make private connectivity free. AWS PrivateLink charges for each provisioned VPC endpoint Availability Zone and for data processed through the endpoint.

For US East and US West (Oregon), AWS's pricing page models a month with one endpoint in one Availability Zone: 720 hours × $0.01 per hour is $7.20, and processing 15.7 GB at $0.01 per GB is $0.16. The PrivateLink portion is $7.36. With the example's $17.50 REST API call charge, the combined total is $24.86. More endpoint zones, traffic, and Region-specific rates change the total. Check AWS PrivateLink pricing for your design.

Free Tier offers depend on account and plan

The API Gateway pricing page lists a service-specific monthly offer of one million REST API calls, one million HTTP API calls, and one million WebSocket messages plus 750,000 connection minutes, for up to 12 months. It also says that, starting July 15, 2025, new AWS customers can receive up to $200 in Free Tier credits.

These account offers have different eligibility rules. AWS's current Free Tier FAQ describes up to $200 in credits for eligible new customers ($100 at signup and up to $100 earned through activities). Its Free plan ends after six months or when credits are exhausted, whichever comes first. On a paid plan, remaining Free Tier credits normally expire 12 months after account creation. AWS says those credits expire immediately if you upgrade by joining an AWS Organization or setting up a Control Tower landing zone. Check the account's credits page for the actual balance and expiry. The FAQ also says customers with an existing or previous AWS account are not eligible for the Free plan or its credits. Because the service pricing page's 12-month API Gateway allowance and the account-level program changed at different times, check the current AWS Free Tier documentation and the account's Free Tier usage page before subtracting any allowance from an estimate.

Service offers and credits do not necessarily cover all costs in an API architecture. For example, the API Gateway page separately describes cache, data transfer, and PrivateLink charges. A budget can alert you when costs approach a chosen amount; it is not a substitute for confirming the account's eligibility or the billable usage.

How to estimate and manage your bill

Choose by required features, then compare rates

HTTP APIs have lower listed request rates than REST APIs, but price alone is not enough to select an API. Compare the features your application requires with AWS's HTTP API and REST API feature comparison. If either type meets the requirements, estimate both using the same request volume, payload sizes, response sizes, Region, and backend assumptions.

Count every billed part

Build the estimate from observed traffic: count HTTP request units from payload sizes, REST calls and outbound bytes, or WebSocket message units and connection minutes. Then add the services that handle each request. A Lambda integration adds separate Lambda request and duration charges; private integrations can add VPC link or load-balancer costs; logs and metrics can add CloudWatch charges. Retries are additional API traffic. If you use API Gateway Portals, include their separate Portal and PortalProducts charges from the current API Gateway price page.

Use the AWS Cost Explorer or a Cost and Usage Report to compare the estimate with billed usage by service, Region, and usage type. For general guidance on checking assumptions and tracing AWS spend, see 10 AWS Cost Optimization Strategies for Engineers. If an integration invokes Lambda, its invocation and duration are separate from the API Gateway charge; AWS Lambda Metrics Explained covers the metrics that help inspect that part of the bill.

Summary

HTTP APIs charge request units in 512 KB payload increments and outbound data. REST APIs charge calls and outbound data, with optional hourly caching and special metering for large streamed responses. WebSocket APIs charge 32 KB message units and connected minutes. Private APIs avoid API Gateway outbound data-transfer charges, but PrivateLink still has endpoint and processing costs. Add backend and monitoring charges, verify Free Tier eligibility for the actual account, and check the current Region-specific price before treating an estimate as a budget.

FAQs

Which API Gateway API type is cheapest?

For the listed request rates, HTTP APIs cost less per request unit than REST APIs. WebSocket pricing uses a different meter: messages plus connection minutes. Choose based on the required features and traffic pattern, then compare total costs with payload, transfer, cache, integration, and Region assumptions included.

Do Private APIs remove all data transfer or networking charges?

No. AWS does not charge API Gateway data transfer out for Private APIs, but AWS PrivateLink can charge for VPC endpoint hours and processed data. Integrations and other services can also add charges.

Does API Gateway caching always reduce cost?

No. REST API caching has an hourly charge. It can save money when the reduction in backend work is worth more than the cache cost, but the result depends on hit rate, response size, and backend pricing.