Caching is straw-man. Complexity tends to be higher to implement a graphql backend than for e.g. type-spec => openapi backend. For the simple case the simpler solution wins imo. For the complex case it can be a toss-up, but it seems that performance and security can be a big deal-breaker for graphql, so maybe only big enterprises can afford to go down that route.

I'm not sure what you mean by the react analogy, react seems to be far more popular than graphql in their respective areas.