A gateway is a proxy that intercepts every request and applies the cross-cutting concerns: who is calling, are they allowed, how much have they used, where does this route, what gets logged. The alternative is implementing all of that in every service, slightly differently, and discovering the differences during an incident.
So yes, once you are past a single API with a single consumer. The value is consistency, not any individual feature.
Two cautions, though. The first is that treating the gateway as “just infrastructure” is how it becomes an ungoverned config sprawl that nobody can audit — dozens of routes, hand-edited policies, and no record of why any of it is the way it is. Drive gateway configuration from your OpenAPI where you can, so the contract and the runtime cannot silently diverge.
The second is that a gateway is a runtime control point, not a design tool. It cannot fix a badly designed contract; it can only mediate one. In a gateway-dominated reality there is a strong pull to treat design-time as optional because “the gateway handles it.” It does not. The consumers built against the contract, and the gateway is downstream of that decision.
