API Gateway Pricing: Models, Cost Drivers, and Examples

An API gateway price is rarely just a software fee. Depending on the product, the invoice may be driven by requests, messages, connection minutes, gateway instances, environments, API proxies, data transfer, analytics, security add-ons, or support. A self-managed open source gateway may have no license fee while still requiring substantial infrastructure and engineering work.

This guide explains the main API gateway pricing models, reviews current public examples from Amazon API Gateway, Google Cloud Apigee, and Kong Konnect, and provides a total-cost worksheet for comparing them with Apache APISIX or API7 Enterprise.

Pricing snapshot: public vendor examples were checked on August 25, 2026. Prices, limits, regions, and packaging change. Use the linked official pricing pages and a written quote for budgeting.

Key Takeaways

Compare the same workload and architecture, not isolated list prices.
Request volume can be important, but environments, transfer, analytics, and operations may dominate the total.
Managed services transfer some operational responsibility to a vendor; self-managed software transfers it to your team.
Open source means the software license can be free, not that production operation has zero cost.
Model expected demand, growth, and failure scenarios before choosing a commercial plan.

The Main API Gateway Pricing Models

Most products use one or more of these models.

Per-Request or Per-Message Pricing

A cloud gateway can meter HTTP API calls, WebSocket messages, connection minutes, or another transaction unit. This model is easy to start with because there may be no fixed license commitment. Cost grows with traffic and can also grow through retries, polling, health checks, or unexpectedly large messages.

Resource or Gateway Pricing

Some platforms charge for a gateway, control plane, runtime unit, CPU capacity, cluster, or environment. This can make the subscription more predictable for steady traffic, but the organization may pay for provisioned headroom that is not used continuously.

Subscription Tier

Enterprise API management platforms often package traffic entitlements, environments, proxies, portals, analytics, governance, and support into a tier. Subscription pricing can simplify procurement but requires careful comparison of included limits and overages.

Customer-Operated Open Source

An open source gateway such as Apache APISIX can be downloaded and operated without a proprietary runtime license. The organization supplies compute, network, observability, security operations, upgrades, and on-call support. Some teams purchase enterprise management or support rather than building every operating capability themselves.

Hybrid Pricing

A vendor may host a management plane while the customer runs data-plane gateways. The invoice can combine a platform subscription or control-plane fee with customer cloud costs. Hybrid is an operating boundary, not a guarantee that either side is cheaper.

Model	Common meters	Main budgeting risk
Per request	Calls, messages, connection time	Spikes, retries, growth, regional rates
Per gateway/resource	Gateways, control planes, CPU, clusters	Idle capacity, topology growth
Subscription	Tier, environment, proxy, entitlement	Limits, overage, renewal changes
Open source self-managed	Infrastructure and labor	Underestimating operations
Hybrid	Platform fee plus customer runtime	Double-counting or omitting responsibilities

Cost Drivers That Apply Across Products

Before reviewing vendors, inventory the dimensions that can affect any gateway deployment.

Traffic Shape

Monthly calls are only one input. Record peak requests per second, payload sizes, response sizes, streaming duration, WebSocket connections, and regional distribution. A million small cached responses and a million large dynamic responses have different infrastructure and transfer costs.

Retries can inflate usage and worsen overload. Model retry attempts at the client, gateway, service mesh, and application. Use bounded backoff and idempotency rules rather than assuming every failed request can be repeated safely.

Deployment Topology

Count production and non-production environments, regions, gateway groups, disaster-recovery capacity, private-network endpoints, and load balancers. High availability normally requires redundant capacity, so a single-instance development estimate is not a production estimate.

Platform Capabilities

Confirm whether these features are included, metered, or restricted to another tier:

authentication and authorization policies;
analytics and retention;
developer portals and published APIs;
audit logs, SSO, and RBAC;
private connectivity and data residency;
custom plugins or transformations;
AI gateway or token-aware controls;
support and availability commitments.

Operations

For every deployment model, assign responsibility for upgrades, patches, backups, certificate rotation, observability, incident response, and capacity planning. Vendor-managed infrastructure can reduce some of that work, while a customer-operated gateway offers more direct control. Neither outcome is automatically better for every team.

Amazon API Gateway Pricing

Amazon API Gateway uses pay-as-you-go meters that vary by API type and region.

This section summarizes AWS as one example inside a cross-vendor pricing-model comparison. For an AWS deployment-specific discussion of infrastructure and operating cost, see Deploying an Open Source API Gateway on AWS.

As of August 25, 2026, the official pricing page states:

HTTP APIs and REST APIs are charged for calls received and data transferred out.
Private REST APIs do not have API Gateway data-transfer-out charges, but AWS PrivateLink charges can apply.
WebSocket APIs are charged by messages sent and received plus connection minutes; message size affects the number of metered messages.
optional REST API caching is charged by cache size and time;
related AWS services such as Lambda, CloudWatch, WAF, and network transfer can add separate charges.

AWS's public examples for selected US regions use $1.00 per million for the first 300 million HTTP API requests and $3.50 per million for the first tier of regional or edge-optimized REST API requests. Those examples are not universal rates for every region or API type. Use the AWS calculator and current regional table for a real estimate.

The AWS page also describes a free tier for eligible new customers. Free-tier rules have changed over time, so do not build a long-term production model around a historical allowance.

AWS Cost Questions

Does the workload need REST API features, or would the lower-cost HTTP API feature set be sufficient?
How much data leaves AWS, and through which network path?
Are Lambda, WAF, logging, or private endpoints required?
How many calls will retries and polling create?
Are response streaming and large payloads metered in additional request units?

AWS is often simple for an AWS-native serverless workload, but the complete cost includes the integrations around the gateway.

Google Cloud Apigee Pricing

Apigee offers both pay-as-you-go and subscription plans. It is a broader API management platform, so its meters include more than gateway requests.

The August 25, 2026 public pay-as-you-go page lists:

Standard API Proxy calls starting at $20 per million up to 50 million calls, with lower rates at higher volumes;
Extensible API Proxy calls starting at $100 per million up to 50 million calls, also with volume tiers;
Base, Intermediate, and Comprehensive environment charges by time and region;
included proxy deployments per environment, with additional deployment units available for Comprehensive environments;
separate prices for API Analytics and Advanced API Security add-ons;
Google Cloud networking charges based on the infrastructure used.

The same official page lists subscription offerings—Standard, Enterprise, and Enterprise Plus—with call entitlements, environments, proxy deployments, analytics retention, and service levels. Subscription prices require contact with Google.

Apigee Cost Questions

Can the API use a Standard proxy, or does it require an Extensible proxy and its policy set?
How many environments, regions, and proxy deployments are needed?
Are analytics and advanced security required for every call?
Is pay-as-you-go or a subscription entitlement more suitable for stable volume?
Which network and hybrid costs sit outside the Apigee line item?

Do not compare Apigee's per-call figure alone with a lightweight gateway. Compare the API management capabilities and environment topology that the organization actually needs.

Kong Konnect Pricing

Kong Konnect combines gateway and usage dimensions. Its public Plus plan on August 25, 2026 lists monthly charges for serverless, hybrid, and dedicated cloud gateway control planes. It includes 1 million API requests per month and lists a charge for additional request volume, with a Plus limit. Dedicated cloud bandwidth, portals, published APIs, analytics, and other modules can introduce additional meters.

Enterprise pricing is custom and billed annually. Fully self-hosted Kong Gateway Enterprise is also described as a separate custom-priced offering. Therefore, older figures based on a previous Konnect rate card should not be reused.

For the current gateway and request figures, see the dedicated Kong Konnect pricing guide. The important comparison dimensions are:

gateway type and number of control planes;
customer-operated data-plane infrastructure for hybrid deployments;
request volume and Plus limits;
dedicated-cloud bandwidth;
portals, published APIs, analytics, and governance;
support tier and an eventual transition to Enterprise.

Apache APISIX and API7 Pricing

Apache APISIX is open source under the Apache License 2.0. There is no proprietary software license fee to download and run the project. A production cost model still includes:

gateway compute and load balancers;
configuration storage or standalone deployment operations, depending on mode;
monitoring, logging, and tracing;
network transfer;
security review, upgrades, and incident response;
engineering time and optional external support.

API7 Enterprise provides enterprise management and support around Apache APISIX. Pricing is quote-based. Ask API7.ai for current commercial terms and scope rather than copying old API7 Cloud figures from a blog or third-party page.

This distinction enables three different operating choices:

self-manage Apache APISIX with internal platform expertise;
use API7 Enterprise to add centralized management and support;
compare those with a fully managed or hybrid vendor service using the same target architecture.

Building a Total Cost of Ownership Model

Use a 12- or 36-month horizon and state every assumption. A defensible model separates subscription, cloud spend, and labor.

TCO = vendor charges
    + gateway infrastructure
    + network and observability
    + operations and support labor
    + migration and implementation
    + risk-adjusted contingency

Step 1: Fix the Workload

Document:

monthly and peak traffic;
average and percentile payload sizes;
protocols and connection duration;
regions and environments;
expected annual growth;
availability and recovery objectives.

Step 2: Fix the Required Capabilities

Separate requirements from preferences. For example, SSO and audit logging may be mandatory in a regulated environment, while a branded external portal may not be needed for an internal API.

Step 3: Price the Same Architecture

If one quote has two production regions and another has one development gateway, the totals are not comparable. Align redundancy, capacity headroom, environments, support, and security.

Step 4: Add Labor Explicitly

Estimate time for platform design, upgrades, on-call, policy management, consumer support, and incident response. Avoid assuming an open source deployment needs a full dedicated team if existing automation can operate it; also avoid assuming it requires no maintenance.

Step 5: Test Sensitivity

Calculate at least three scenarios:

Scenario	What changes
Expected	Forecast traffic and normal operations
Growth	Higher traffic, APIs, regions, and retention
Stress	Retry amplification, failover transfer, temporary capacity

Also identify thresholds that force a plan change or enterprise negotiation. A sharp tier transition can matter more than small differences at the starting volume.

Example Comparison Worksheet

Category	Managed pay-per-use	Enterprise subscription	Open source self-managed
Request or entitlement fee	Vendor meter	Included/overage terms	None from the project
Gateway runtime	Often vendor-operated	Vendor- or customer-operated	Customer-operated
Environments and regions	Service-specific charges	Tier or contract	Customer infrastructure
Analytics	Included or metered	Tier/add-on	Customer tooling
Security/governance	Service or add-on	Tier/contract	Gateway plus customer controls
Support	Cloud support plan	Contract	Community or paid support
Upgrades	Mostly vendor service	Shared responsibility	Customer responsibility
Portability	Product-specific	Product-specific	Depends on configuration and extensions

This table does not declare a winner. It shows where to place cost and responsibility.

Common Pricing Mistakes

Comparing Only the Cheapest Request Rate

A per-million rate may exclude environments, transfer, analytics, or the features required by the workload. Start from the architecture and billable units, not the smallest number on the page.

Treating Throttling as an Overage Mechanism

A gateway's quota or throttling behavior may reject traffic rather than simply billing it. Some quotas are best-effort and should not be used as a financial control. Use provider budgets and monitoring as well as gateway limits.

Assuming Open Source Has No Cost

The license can be free while reliable production operation is not. Include real infrastructure and team effort.

Treating Every Vendor Feature as Required

Buying a broad enterprise tier for unused modules can cost more than operating the necessary subset. Conversely, excluding required governance to make a smaller plan look cheaper creates an invalid comparison.

Using Undated Blog Prices

Vendor packaging changes. Date every snapshot, preserve the official URL, and refresh figures before approval and renewal.

Frequently Asked Questions

How is API gateway pricing calculated?

Common units include calls, messages, connection minutes, data transfer, gateway resources, environments, API proxies, subscription tiers, and support. A product may combine several units.

Why can API gateway cost rise at scale?

Traffic growth, retries, larger payloads, more regions, longer retention, advanced security, and higher support requirements can all increase cost. Growth can also cross a plan limit and require a different contract.

Is an open source API gateway free?

The software may be free to use under its license. Production infrastructure, observability, security, upgrades, and operations still cost money.

Which pricing model is best?

There is no universal answer. Pay-per-use can suit variable or early workloads. A subscription can suit a governed enterprise program. Self-managed open source can suit a team that values runtime control and already has platform automation. Evaluate total cost and responsibility for your workload.

Conclusion

API gateway pricing is an architecture and operations question expressed in commercial units. Amazon API Gateway, Apigee, Kong Konnect, Apache APISIX, and API7 Enterprise distribute cost and responsibility differently.

Build a fixed workload, identify required capabilities, price the same topology, and include infrastructure and labor. Then refresh public rates and request written quotes. That process produces a more reliable decision than any table that compares one request price while ignoring the system around it.