Every technology has some degree of lock-in. Serverless has tradeoffs like anything else. The question is, do the benefits outweigh the costs? I recently did a writeup on this that helped me work through my own thoughts on the issue: https://www.trek10.com/blog/think-faas-podcast-talkin-lock-i...
"lock-in" potential should be measured in how difficult it is to escape a platform/system/framework.
E.g., if I become unhappy with k8s on Google Cloud, it's easy to migrate to aws. If I become unhappy with k8s, it's easy to migrate to Mesos. If I become unhappy with Linux, it's easy to migrate to FreeBSD. If I'm unhappy with Lambda, it's easy to migrate to cloud functions, or a self-hosted FaaS.
Some systems require major rework, or are just otherwise impossible to migrate off (probably by design). Those are the systems to be careful when adapting.
Evaluate the "exit friction", build something useful, and stop worrying.
In the late 90s to early 2000s we called this "Application Service Providers" or "Managed Hosting."