"it's ORMs that are only good for fairly simple data models" That's not true. Maybe you haven't used a powerful ORM, but with a good ORM you can scrape a fair

"it's ORMs that are only good for fairly simple data models" That's not true. Maybe you haven't used a powerful ORM, but with a good ORM you can scrape a fair bit of bad architecture under the rug. "SQL is good for both simple and overly complex models" SQL is poor for complex models. Imagine a schema refactor, and then a SQL query that breaks on aggregation without you getting any notice of it. It's easily done. Maintenance is a key quality of good software. There's other maintenance problems with platforms based around reldbs as well: it's arduous to maintain parallel test and production environments, and to have dev environments that operate at the same scale as your production environment.