When you try to understand a new thing such as GraphQL, it helps to compare it with something existing, for example, REST, which you already know.
Imagine we have several web and mobile applications that retrieve data from the same server. In RESTful architecture, we design each entity as a resource. When request for fetching a resource is received, the server usually returns everything about that resource. Thus, the clients get redundant and unnecessary data which consumes bandwidth. Depending on the scenario, this can total an amount significant enough for the client's performance (think about mobile clients).
How about the clients specifying exactly which data they need and the server sends only those data? GraphQL enables us to achieve this.
Like pretty much everything in life, it depends. Not all projects, regardless of their sizes, have the same requirements. Determine your project's requirements. Consider the available technologies and their pros and cons. It's a trade-off. There's no silver bullet or one-size fits all solution. Nonetheless, Facebook uses GraphQL and there are strong reasons to consider their project as BIG.
