Lambda@Edge is primarily event driven for requests, yes the latency is reduced but it is designed to be used with an origin. Even if you have the Lambda@Edge function your CloudFront distribution would still need to reach out to the origin.

When the origin is an API Gateway (with or without a Lambda) there are so many benefits come with it such as:

• Built in support for OpenAPI spec
• Built in routing
• Support for canary deployments
• Authentication and authorization built in, with support for throttling per API user
• Built in support for model schema validation
• Support to manage your API as Infrastructure as Code: CloudFormation, Terraform, SAM
• Each request can be developed separately, you can only have 1 Lambda@Edge function per event so you might see collisions in a team (not to mention very large functions).

If this was all needed to be done by Lambda@Edge then you would need to design, develop and maintain these features.

The price you refer to is for REST API, bare in mind there 3 kinds of API the cheapest is HTTP API which costs $1.00 per million requests.

Lambda@Edge has many limitations as compared to regular lambda. For examples:

• its only for node.js and python,
• no lambda layers!
• no AWS X-Ray!
• max timeout 30 seconds,
• and many more.

Thus, it would be difficult in many cases to migrate to Lambda@Edge and to give up all these benefits of regular lambda.