Your question is way too broad to answer it in few sentences... Normally what I'd do is analyse the project as much as possible (goals, planned growth, market, etc) then build schema with each components challenged with few similar services. Then re-analyse it again from financial perspective and possible technical debt. Then ask to review it someone else to avoid my biases holding me from optimal solution.
From what I see nowadays, it's more like, hey we have a project... Let's do it with AWS it has a ton of cool features!
(substitute AWS with GC, Azure or anything of your choice)