Hah! Hey, well, you're right... It seems I'm also not immune to AWS propaganda. Regardless, my original point stands: Serverless doesn't mean infinitely-scalable. There are machines somewhere running those containers, and if your traffic spike was sufficient to be notable on a global scale, there are going to be limits like with any other technology. Outsourcing the containers and everything underneath doesn't automatically give your application more capacity to scale. Just to use the example of Lambda, at some point, you will hit a capacity whereafter AWS will be spinning up EC2 instances to accommodate their containers that run your functions, and those take some minutes to spin up. Now, all that being said, I don't pretend to have any special insight into the way AWS builds their reserve infrastructure, though I'd be very interested to have that insight. Just through a layer of abstraction and barring any special preparation of reserve capacity, serverless won't automatically give you more capacity to scale than running your own containers or EC2 instances or bare metal somewhere.