@Ighor Martins Assuming you most of your queries are read, you should check out the caching service presented yesterday - graphcdn.io

I'm not sure to understand. Let's say I have a medium sized app (50/100 rps), mostly read queries. I would total between 4-8M requests/day. At $10 / 1M requests, it would be insanely expensive, totally not worth it, knowing that redis/memcache could run at 60-200k rps on server costing a few dollars/month.Am I missing something? Or this service is more for big companies with big $ wanting an easy solution to scale their graphql api?

Not exactly sure how you would be able to get that redis-server to run for a few (<$10 at least) per month, but generally what a service like graphcdn gives (we don't use it yet, but mostly due to other considerations than price): 1. Edge caching - Responses to clients, if cache hits, will be very fast - this has a value in itself 2. The caching logic, including invalidation, which for GraphQL can be complex to get right (and also something you would have to pay for if you were to roll it yourself). I think that for any somewhat complex graphql backend, 1M requests will cost quite a bit more than $10, hitting maybe several internal endpoints. I am not an expert at these things, but if you think that service is insanely cheap, and can prove it with your own, you have a real business propsotion 🙂