AWS API Gateway vs Kong Konnect vs Tyk Cloud: Real API Gateway Cost per 100 Million Requests in 2026
The short answer: at 100 million requests a month, only AWS API Gateway publishes a self-serve rate that actually extends to this volume, and it is cheap: about $186 a month on the HTTP API type (requests plus data transfer), or about $436 on the pricier REST API type. Kong Konnect's self-serve Plus plan is explicitly capped at 10 million requests a month — at that cap, Plus costs about $1,825, but 100 million requests is ten times past where Plus stops selling at all, and Kong's own pricing puts anything beyond the cap into Enterprise, custom-quoted territory. Tyk Cloud's public pricing does not disclose a dollar figure precise enough to compute a real monthly total at 100 million requests: its Professional tier is flat-rate and unlimited-request, but Tyk does not publish what that flat rate is, so Tyk's cost at this workload is QUOTE-ONLY / UNKNOWN. The honest comparison at genuine production scale is AWS's published number against two custom quotes, not three comparable self-serve prices.
Kong and Tyk are not sold as pay-per-request products at this volume
Both vendors' entry-level cloud offerings are explicitly bounded. Kong Konnect's Plus plan states a 10 million requests/month maximum on its own pricing page — it is not a product designed to be extrapolated past that ceiling, and Kong's own materials route anything beyond it to Enterprise. Tyk's Professional tier is flat-rate with no consumption limit, which is the opposite problem for this article's purpose: it removes any per-request calculation entirely, and Tyk does not publish what that flat rate costs. This article treats both vendors' stated limits as a line that should not be crossed in either direction — not extrapolated past Kong's cap, and not invented for Tyk's undisclosed Professional price.
What each vendor actually bills
AWS API Gateway (OFFICIAL, aws.amazon.com/api-gateway/pricing). HTTP APIs: $1.00 per million requests for the first 300 million a month, stepping down to about $0.90 per million above that. REST APIs: $3.50 per million requests for the first 333 million a month. WebSocket APIs: $1.00 per million messages plus $0.25 per million connection-minutes. Data transfer out: $0.09 per GB at the first tier, on AWS's standard tiered egress schedule. Free tier: 1 million REST, 1 million HTTP and 1 million WebSocket calls a month for a new account's first 12 months. No separate platform or per-gateway fee exists on either API type — cost is pure usage, and the rate applies identically from the first request to hundreds of millions.
Kong Konnect (OFFICIAL, konghq.com/pricing). Plus: $25/month per serverless control plane (hybrid control planes run $200/month, dedicated cloud control planes $500/month plus $0.15/GB bandwidth). 1 million API requests/month included, then $200 per additional 1 million — up to a stated maximum of 10 million requests/month. At that 10-million-request ceiling, Plus costs about $1,825/month ($25 base + 9 million overage requests at $200/million). Above 10 million requests, Plus is not available as a self-serve product — Kong's own pricing directs higher-volume customers to Enterprise, which is fully custom and requires a commercial conversation. Developer portals and published-API counts are separately metered add-ons on Plus, not modeled here.
Tyk Cloud (OFFICIAL structure; no disclosed dollar figure for the relevant tier). Core: usage-based, for Cloud, Self-managed and Hybrid deployments — Tyk's own site describes this as flexible consumption pricing but does not publish a per-request dollar rate. Professional: flat-rate, with no consumption limits — unlimited API gateways, unlimited APIs and unlimited requests a month — but Tyk does not disclose what that flat rate costs. Enterprise: custom, for high-traffic, multi-cloud and multi-region environments. Tyk also ships a free, open-source self-hosted gateway independent of the managed Cloud product, with no license fee but real infrastructure and operational cost to run at scale.
Normalizing the workload
All ILLUSTRATIVE. 100 million API requests a month; an average combined request-and-response payload of 10 KB; one production gateway; moderate global traffic; no LLM inference cost. That payload assumption gives about 954 GB of data transfer a month, priced at AWS's $0.09/GB first-tier rate. Kong's published rate card does not break out a separate data-transfer charge from its per-request pricing.
Cost at the normalized workload
Formula: AWS = (requests ÷ 1,000,000 × per-million rate) + (GB transferred × $0.09); Kong Plus = $25 + ((requests − 1,000,000) ÷ 1,000,000 × $200), valid only up to the stated 10-million-request cap; Tyk = QUOTE-ONLY / UNKNOWN at any volume, since Professional's flat rate is not published.
| Platform / plan | Basis | Monthly cost |
|---|---|---|
| AWS API Gateway, HTTP API | 100M ÷ 1M × $1.00 + $86 data transfer | $186, at 100 million requests |
| AWS API Gateway, REST API | 100M ÷ 1M × $3.50 + $86 data transfer | $436, at 100 million requests |
| Kong Konnect Plus, at its published 10M cap | $25 base + 9M ÷ 1M × $200 | $1,825, but only covers 10 million requests — 10% of this article's workload |
| Kong Konnect, above 10M requests | Not sold as self-serve Plus; routes to Enterprise | QUOTE-ONLY, for the 100-million-request workload |
| Tyk Cloud, any tier covering 100M requests | Professional is flat-rate and unlimited-request, but the rate is not published | QUOTE-ONLY / UNKNOWN |
AWS's HTTP API type is the only line in this table with a confirmed number at the actual 100-million-request workload. Kong's $1,825 figure is a real, calculable price, but it is the price of 10 million requests, not 100 million — a tenfold gap between what Plus is designed to sell and what this article is pricing.
Why Kong's cap matters more than its rate
Formula: Kong Plus cost at its cap = $25 + 9,000,000 ÷ 1,000,000 × $200 = $1,825, for 10 million requests — an effective rate of about $182.50 per million, well above AWS's $1.00–$3.50 per million. Even at the cap, Kong Plus is far pricier per request than AWS, reflecting that it is sold as a managed gateway platform with plugin ecosystem, developer portal and hybrid/dedicated deployment options, not a bare pay-per-request product. The more important fact for this article's specific workload is that Kong does not let a self-serve customer buy past 10 million requests a month on Plus at all — the product itself enforces the boundary, which is why extrapolating its per-request rate tenfold to reach 100 million would describe a product Kong does not sell.
Why Tyk cannot be priced here
Tyk's Professional tier is explicitly structured to avoid a per-request calculation: it is a flat monthly fee for unlimited requests, which is the right pricing shape for a high-volume, predictable workload, but only if the flat fee itself is known. Tyk's public pricing page does not disclose that figure. Any dollar estimate for Tyk at 100 million requests a month would be invented, not confirmed, so this article marks it QUOTE-ONLY / UNKNOWN rather than supplying one.
Sensitivity
- Whether traffic is genuinely self-serve-sized or requires an Enterprise conversation. The single largest lever for Kong and Tyk: 100 million requests a month exceeds Kong's published self-serve ceiling entirely and sits well above what either vendor's publicly priced tier is built for.
- HTTP versus REST API type, on AWS specifically. A 3.5x price difference for the same request volume, trading away REST's caching and advanced usage-plan features.
- Average payload size and data transfer. AWS's data-transfer charge scales with it directly; neither Kong's nor Tyk's published materials separate out a comparable data-transfer line.
- Per-gateway-service structure, on Kong specifically. Splitting traffic across multiple gateway services each incurs its own base fee, which changes the effective blended rate depending on how traffic is partitioned, within the 10-million-request ceiling.
Budgeting traps
- Extrapolating Kong's $200-per-million overage rate past its published 10-million-request cap. Kong's own pricing does not sell Plus beyond that point; a literal tenfold extrapolation describes a product that does not exist.
- Inventing a Tyk Professional dollar figure from its "flat-rate, unlimited" description. The description tells you the pricing shape, not the price; Tyk does not publish the number.
- Comparing AWS's per-request price to Kong's or Tyk's without accounting for feature differences. Kong and Tyk bundle plugin ecosystems, on-premises or hybrid deployment options, and developer portals that AWS API Gateway does not offer in the same form.
- Choosing REST API on AWS without needing its specific features. HTTP API is 71% cheaper for workloads that don't need REST-specific caching, request validation or WAF integration at the gateway layer.
- Forgetting downstream costs. None of these figures include the compute (Lambda, containers) that actually serves the request, which is typically several times the gateway's own cost.
What to ask before you buy
Ask Kong directly for Enterprise pricing once your traffic exceeds the published 10-million-request Plus ceiling, since no higher self-serve tier exists to extrapolate from. Ask Tyk directly for its current Professional flat-rate price at your expected volume, since the public pricing page states the pricing shape but not the number. Confirm whether HTTP API's feature set meets your needs before defaulting to the pricier AWS REST API type, and model your downstream compute cost alongside any of these gateway figures.