So far I've only used Next.js for static site generation, for a couple of long-term projects. Self-hosting it is as easy as any static site. Upgrading major versions of Next.js, however, has been fairly painful. The experience made me reconsider the decision to use it, and I'm keeping up on possible replacements like Remix. But honestly I'm getting a lot of value out of the framework that I'm in no hurry to change.

For dynamic sites that require running Next.js on the production server, I'm not too interested in trying because it feels like too much vendor lock-in. The same reason I wouldn't consider Vercel for hosting, since they develop the framework that is already a big dependency.

I think this is the key though really, and what peeves me the most.

NextJS is fine to self host if you don't need any of the features that actually make it worth dealing with all the additional complexity.

Essentially if you want any of the special sauce you have to host on vercel, or use opennext guides to build equivalent infra.

Its just not worth the complexity for me to use.