Billing Dimensions in Confluent Cloud

Dimension summary

Clusters are billed based on the dimensions listed in the following tables. For every available dimension, the following table lists the Costs API line item and the unit of measure for the dimension.

Basic, Standard, Enterprise, and Freight Kafka clusters

The following table summarizes the billing dimensions for Basic, Standard, Enterprise, and Freight clusters. Capacity is measured per eCKU. The following sections contain more details for some of the billing dimensions.

Dedicated clusters

The following table summarizes the billing dimensions for Dedicated clusters. The following sections contain more details for some of the billing dimensions.

CKU

A CKU is a fixed unit of pre-allocated capacity for Dedicated Kafka clusters in Confluent Cloud. Each CKU bundles throughput, partition, and connection limits. You set the number of CKUs at cluster creation and pay per CKU per hour.

CKU charges accrue each hour based on the configured capacity of your cluster. You can expand and shrink Dedicated clusters. When you expand or shrink a Dedicated cluster, you are billed for the new CKU value in the next hour following the change in capacity.

eCKU

An eCKU is a unit of capacity for Basic, Standard, Enterprise, and Freight Kafka clusters in Confluent Cloud. Unlike CKUs, eCKUs scale elastically: you pay only for the capacity you use in a given hour, up to a fixed ceiling.

Basic, Standard, Enterprise, and Freight clusters are elastic and automatically scale up to a fixed ceiling. The ceiling governs how much maximum capacity your Kafka cluster can use; however, you are only billed for the capacity that you use in a given hour, up to the maximum allowed by the ceiling. To determine the number of eCKUs used in a given hour, the billing system monitors the actual consumption of each dimension.

Basic, Standard, Enterprise, and Freight clusters have a minimum capacity if you use any capacity. If the capacity you use is less than the minimum, you pay the minimum. If your cluster is at zero consumption across all dimensions, you pay nothing. Zero consumption across all dimensions means your cluster has no partitions, no topics have been created (or all topics have been deleted), and there is no other capacity usage from an eCKU billable dimension, including the following: ingress, egress, client connections, connection attempts, or requests. For more information, see eCKU/CKU comparison and Cluster provisioning and scaling.

Compare billing units for Kafka clusters

CKUs and eCKUs are billing units for different Kafka cluster types. Both quantify pre-allocated resources; they differ in how you are billed.

Ingress and egress

Confluent Cloud charges you for all data transfers, including requests for produce and consume and every other type of incoming or outgoing request. This includes charges for requests made by Cloud Console to display messages in message browser. There is an additional $0.01 ingress charge for using the Produce (v3) REST API, before the regional multiplier is applied. All ingress and egress charges are then multiplied by a regional multiplier based on the cloud provider region.

Storage

Confluent Cloud charges for storage based on the total post-replication volume of data on a Kafka cluster. Because Confluent Cloud replicates data three times for high availability, billed storage is typically about 3x the volume you write. To reduce storage usage, you can configure the retention.bytes and retention.ms properties on a topic. To learn more, see Topic Settings and Edit topics.

Partitions

Confluent Cloud does not charge for partitions on any type of Kafka cluster, but the number of partitions you use can have an impact on eCKU. To determine eCKU limits for partitions, Confluent Cloud bills only for pre-replication (leader partitions) across a cluster. For more information, see eCKU/CKU comparison.

Cluster Linking

Cluster Linking in Confluent Cloud is billed on three dimensions: cost per cluster link per hour, ingress per GB written, and egress per GB read. A cluster link connects a source cluster to a destination cluster. For more details, see Supported cluster types for Cluster Linking.

The following tables provide a detailed breakdown for Cluster Linking billing.

Examples

The following examples illustrate cluster-linking billing for common configurations. These examples are for clusters with internet networking only; examples of private networked clusters are provided in later versions of this documentation.

One cluster link in an active-passive setup

In this example, you have one cluster link between a Source cluster and a Destination cluster.

The source cluster is charged $0.25 per hour and $0.25 per GB.

The destination cluster is charged $0.25 per hour and $0.01 per GB, because the cluster’s KafkaNetworkWrite cost is $0.01 per GB.

Two cluster links in an active-active setup

In this example, you have two clusters with two cluster links between them: one in each direction.

Each cluster is charged $0.50 per hour (2 x $0.25).

Each cluster is charged $0.25 per GB that it sends over a cluster link.

The left cluster is charged $0.04 per GB for data it receives on a cluster link, because that is its KafkaNetworkWrite price.

The right cluster is charged $0.01 per GB for data it receives on a cluster link, because that is its KafkaNetworkWrite price.

Hybrid cloud or migration with one link writing to a Confluent Cloud cluster

In this example, you have an external source cluster, which is not a Confluent Cloud cluster, with one cluster link sending data to a Confluent Cloud cluster.

The Confluent Cloud cluster is charged $0.25 per hour and $0.01 per GB, because that is its KafkaNetworkWrite price.

The external cluster does not accrue Confluent Cloud charges. If the external cluster is hosted in a public cloud provider, it may be charged an egress fee to send data to Confluent Cloud.

Managed connectors and custom connectors

Review the following billing information for managed connectors and custom connectors. For both managed and custom connectors, throughput billing is based on decompressed data. This is the metric reported by the connector. In Cloud Console, you can see decompressed data metrics on the tile in Connectors. The Kafka cluster reports compressed data. In Cloud Console, you can see this (lower) metric on the cluster tile in Environments.

Managed Connect cluster

If you use a dedicated Kafka cluster, connectors on Confluent Cloud run on your own dedicated Connect cluster. Confluent Cloud provisions the dedicated Connect cluster when you launch your first connector. If you use a Basic or Standard Kafka cluster, your connector runs in multitenant Kafka Connect clusters. For the connectors running on a multitenant Connect cluster, there is no Connect capacity (ConnectCapacity) charge.

If you pause a managed connector, tasks are paused, which means message processing is also paused until the connector is restarted. However, the connector is still active and hourly base costs for tasks assigned to the connector continue to accrue. To stop your Connect connectors from accruing billed usage, you can delete them in the Confluent Cloud Console, or use the Confluent CLI.

The dedicated cluster and fully-managed connectors are billed based on the following dimensions. For Custom Connector billing details, see Custom connector cluster.

Dimension	Unit of measure
Dedicated cluster	Cost per hour
Task base price	Cost per task per hour
Throughput	Cost per GB written or read per hour

Fully-managed connector pricing is displayed in the Add Connector screen:

• Dedicated connect cluster hourly pricing is shown on the Review and launch screen the first time you launch a connector.

• Dedicated cluster billing is shown as a line item on the Billing and payment > Billing screen.

• Self-managed connectors have no billing mechanism themselves. However, note that using self-managed connectors may incur ingress, egress, and storage charges for your Kafka clusters running in Confluent Cloud.

Custom connector cluster

Custom connectors are supported on Basic, Standard, and Dedicated clusters. Your custom cluster is billed based on the following dimensions:

Dimension	Unit of measure
Task base price	Cost per task per hour
Throughput	Cost per GB written or read per hour

Note

• Customers are responsible for all charges related to using the app log topic with a custom connector.

• Customers incur some throughput charges while using APIs, CLIs, and the UI to access logs data stored in the apps log topic.

• There is a 750 Mbps throughput limit.

ksqlDB

Fully-managed ksqlDB is billed based on the following dimension:

In addition to the per-CSU charge, ksqlDB applications may influence Kafka ingress, egress, and storage.

Fully-managed ksqlDB pricing is displayed in the Add an Application screen:

To stop your ksqlDB application from accruing billing, you can delete your application in the Confluent Cloud Console, or use the Confluent CLI to delete your application.

Connector events and audit log

Connector events and audit logging operate from the Audit Log cluster. Consumption charges for both connector events and audit logging are combined on your bill.

Flink SQL

Cloud native Confluent Cloud for Apache Flink is billed based on the following dimension:

With the serverless architecture of Confluent Cloud for Flink SQL, you’re charged only for the minutes when your queries are running.

For more information, see CFU billing.

Tableflow

Tableflow is billed by topic-hour and gigabytes (GB) processed.

Billing is enabled at a per-topic level, which includes Apache Iceberg™ and Delta Lake table formats. If you enable a topic for both formats, you’re only charged once for the overall topic, not per format.

For more information, see Billing with Tableflow in Confluent Cloud.