If enough people start using it? I'd suggest doing the dynamic part on a tiny instance (i.e. only as big as you need for scraping and generating the updates / trends) and pushing updates to S3. Then the client can poll S3 objects. (With some marker for "latest")

You'd have to cross 1gb/h of transfer before it costs more than running more than t3.small.

That's a great hint thanks. I just created a test bucket with a JSON file, this seems much cleaner. At the moment I still use a node server and the database on the same instance.