You don't mention the total cost of ownership (TCO) in your article. Speaking only about infrastructure is a mistake many people are doing. The EC2 instance needs to be updated/patched. You generally need people to do this, and the cost of an ops/engineer is much more than hundreds of thousands of Lambda invocations. Also you cannot strictly compare one EC2 instance with one function. Functions are spread over multiple AZs with built-in load balancing. You would need a load balancer with at least 2 instances (on different AZs) to achieve the same availability... I agree with you it can sometime be more cost effective to use EC2/Fargate but do not neglect the cost of the human workforce needed to manage it when comparing.