API gateway pricing compared across four clouds

Por Rohit Athaide · Publicado el 4 nov 2026

Short answer: an API gateway sits in front of your services to route calls, check keys, limit rates and collect logs. It is billed per request or per unit of capacity per hour, and the two models suit different traffic. Per-request billing costs almost nothing when idle. Capacity billing costs more at rest and less per call at high volume.

Key points

• Two billing models: per request, and per hour or tier.
• Compare a lightweight gateway to a lightweight gateway, and full management to full management.
• Caching, logging and data out are billed on top.
• A function-based API normally needs a gateway.
• Check the free allowance before sizing.

What are the two billing models?

1. Per request. You pay for each call, quoted per million. Cost follows traffic and is near zero when idle. Good for small or spiky APIs.
2. Per hour or per unit. You pay for gateway capacity whether or not it is used, and often get lower or no per-request charges. Good for steady high traffic and for features such as private networking.

Some clouds sell both models as separate tiers. Check which tier a comparison uses before you trust a total.

Which services are equivalent?

A lightweight gateway and a full management product solve different problems. A comparison that sets one against the other is not fair. Our API gateway compare page shows which product and price lines we match for each cloud.

What does a worked example look like?

Made-up round numbers to show the method, not real prices. An API handles 30 million calls a month, with an average response of 20 KB.

At 30 million calls the per-request model is lower. At 150 million calls the per-request line becomes 525.00 while the capacity line stays near 219.00, so the order flips. The break-even is where per-request cost equals the hourly price, here about 62 million calls. Plot your expected volume against it.

What changes the total?

• Request count and size. Some meters round up by message size.
• Caching. A cache cuts backend calls but is billed by size and hours.
• Logging. Access logs are billed as ingested data. See logging and monitoring costs.
• Data out. Responses leaving the cloud are egress. See data transfer costs.
• Authorizers and custom code that run on every call.

Do I need a gateway at all?

If one load balancer already sends traffic to your servers, a gateway adds cost for features you may not use. A function-based API often needs one, since the gateway is what exposes the function to the internet. See Lambda vs Azure Functions vs Cloud Functions and load balancer costs.

Which mistakes are common?

• Using a full management tier for a three-endpoint API.
• Leaving full request logging on in production.
• Buying capacity for a peak that happens once a month.
• Calling the gateway for internal service-to-service traffic.

Decision checklist

• Is volume below or above the break-even for per-request billing?
• Do I need keys, quotas and a portal, or just routing?
• Is the cache sized for the hot paths only?
• Are logs sampled?

Next steps

The API gateway compare page shows each price line with its unit and a link to the vendor’s page. The serverless estimator includes the gateway in a full function workload.