GraphQL is an interesting one. I, personally, don’t really like GraphQL and serverless, mainly because it tends towards misuse of data stores and allows a service user to add unexpected loads onto a serverless scenario which could produce bottlenecks. That being said, it’s a really useful shortcut for a lot of simple scenarios. However, those simple scenarios are really the exception in the majority of systems I’ve seen. It’s why GraphQL is often great for prototyping an initial solution but not necessarily great longer term.
Please note: personal opinion!!!