NoSQL is great as long as you know all the query patterns and data validity constraints up front. Schema migration, foreign keys, and data validation generally turn out to be a lot harder in NoSQL solutions than traditional SQL, so it entirely depends on your use case.
The platform that will stay the most constant is "separate services using software frameworks that are popular and/or you control" talking to "back-end services that uses popular technologies."
Spring + Hibernate + MySQL isn't going to go away anytime soon. Redis, Postgres, and Mongodb also seem to be around for a long time (although Mongodb has serious operational challenges! You can't even take a consistent backup easily.) Host on AWS if you want stability. Azure and Google Cloud have less of the market, and Google has been known to just give up on products and shut them down at times, but they seem really tenacious about their Cloud server / hosted-kubernetes type options, so that's probably safe.
Google has BigTable, Spanner, or Firestore and Amazon has DynamoDB or Neptune if you want a hosted, NoSQL database that scales very large.
An alternative is to use something like Google Firebase, or Amazon AppSync, which is more of a "ready back-end" -- as long as you don't need millisecond-latency simulation.