Data Plane · Streaming & Stream Processing

Confluent

Data streaming platform built around Apache Kafka, available as a cloud service or self-managed.

Overview

Confluent is not a competing protocol to Kafka. It is the commercial platform and managed service built around it, by the people who created Kafka at LinkedIn, and it should be compared with managed Kafka rather than with Kafka itself.

It comes in two main shapes: Confluent Platform, the self-managed software, and Confluent Cloud, the fully managed service, alongside Confluent Private Cloud and WarpStream.

Version alignment is straightforward. Confluent Platform 8.3 ships Apache Kafka 4.3, so you are not left behind the open-source release train.

What it adds around the log is the actual product: Schema Registry, REST Proxy, ksqlDB, managed connectors, Stream Governance, role-based access control and managed Apache Flink for stream processing.

Licensing needs care in a procurement review. Kafka and the clients stay Apache 2.0, but Schema Registry, REST Proxy, ksqlDB and community connectors are under the Confluent Community License, which allows embedding and redistribution but forbids offering a competing service.

The current direction is the lakehouse bridge. Tableflow materialises Kafka topics as Iceberg or Delta Lake tables, handling schemas, type conversion, evolution and catalogue publishing, with freshness of roughly five minutes.

Those tables publish into AWS Glue, Databricks Unity Catalog, Apache Polaris or Snowflake Open Catalog, with bring your own or Confluent-managed storage.

The second direction is AI, marketed as Confluent Intelligence and Streaming Agents, built on the same managed Flink layer.

Ownership changed materially in 2026: IBM completed its acquisition of Confluent in March 2026, so any description of it as an independent Nasdaq-listed company is now historical.

On pricing it is the most transparent of the commercial streaming vendors, publishing cluster and Flink rates, though connector, Tableflow and Stream Governance rates and all dedicated-cluster pricing are withheld.

Key features and capabilities

The same headings are used for every streaming & stream processing entry, so two tools can be read side by side.

What it does
  • Kafka's log model unchanged, with topics, partitions and consumer groups
  • Cloud clusters come as Basic, Standard, Enterprise, Freight and Dedicated, billed in elastic capacity units
  • Dedicated clusters use pre-provisioned capacity rather than elastic units
  • Topics can be materialised as Iceberg or Delta tables through Tableflow
Delivery and ordering guarantees
  • Inherits Kafka's per-partition ordering, idempotent producer, transactions and exactly-once semantics
  • Platform 8.3 ships Kafka 4.3, including the server-side consumer rebalance protocol
  • Managed Flink adds checkpointed exactly-once state on top of the log
  • Delivery guarantees are Kafka's, so migrating in or out does not change application semantics
Scale and retention
  • Throughput is published as bands per cluster type rather than partition counts
  • Freight clusters require a minimum of two elastic capacity units
  • Storage is billed per GB a month, so retention is elastic rather than capped
  • Tableflow extends retention indefinitely into Iceberg or Delta tables in object storage
Processing and transformation
  • Confluent Cloud for Apache Flink covers SQL, Table and DataStream work, billed per Flink unit
  • Kafka Streams, Apache 2.0, and ksqlDB, Confluent Community License, ship with the platform
  • Streaming Agents and Confluent Intelligence target AI-driven processing
  • Processing is a first-party managed service rather than something you operate separately
Security and governance
  • Local accounts, single sign-on, service accounts, API keys and OAuth or OIDC identity providers
  • Role-based access control, ACLs and IP filtering
  • TLS in transit, self-managed encryption keys on all three clouds, and client-side field-level encryption
  • PrivateLink, VPC peering, transit gateways, network isolation and custom domains, with activity monitoring
  • Certifications include SOC 1, 2 and 3, ISO 27001 and 27701, PCI DSS, HITRUST and HIPAA, plus a government cloud
Connectors and ecosystem
  • Managed and self-managed connectors through Confluent Hub, though no total count is published
  • Full Kafka protocol and client compatibility
  • Schema Registry supporting Avro, Protobuf and JSON Schema
  • Lake and warehouse reach through Tableflow into Glue, Unity Catalog, Polaris and Snowflake Open Catalog
Where it runs and what it costs
  • Self-managed Confluent Platform 8.3, or fully managed Confluent Cloud
  • Cloud on AWS, Azure and Google Cloud, covering roughly 30 AWS, 28 Azure and 26 Google Cloud regions
  • Also AWS GovCloud and Jio Cloud regions, with unlisted regions available on request
  • Confluent Private Cloud and WarpStream cover the in-your-own-infrastructure cases

Pricing

Usage-basedBasic from $0 a month

Basic gives the first elastic capacity unit free, then $0.14 an hour with data at $0.05 a GB and storage at $0.08 a GB a month, starting effectively at zero. Standard is $0.75 a unit hour, roughly $385 a month; Enterprise $1.75 to $2.25, around $895; Freight $2.25 with cheaper data and storage, around $2,300. Managed Flink is $0.21 a unit hour, metered per minute. Connector, Tableflow, Stream Governance and dedicated-cluster rates are not published. New signups get $400 to spend in 30 days.

Vendor pricing page →

Demos and videos

About Confluent, an IBM company

Confluent was founded in 2014 by Jay Kreps, Jun Rao and Neha Narkhede, the co-creators of Kafka at LinkedIn, and is based in Mountain View with a London office. It listed on Nasdaq as CFLT in June 2021 at $36 a share. IBM completed its acquisition on 17 March 2026, buying all outstanding shares at $31 each for an enterprise value of roughly $11bn, so it is now an IBM company rather than an independent public one. Licensing is split three ways: Apache 2.0 for Kafka and clients, the Confluent Community License for Schema Registry, REST Proxy and ksqlDB, and commercial terms for the cloud and enterprise features.

Founded 2014 · Mountain View, California · confluent.io

Other streaming & stream processing tools

Amazon Kinesis

Data Plane · Streaming & Stream Processing

AWS services for collecting and processing real-time streaming data.

  • Cloud service

Apache Flink

Data Plane · Streaming & Stream Processing

Stateful stream processing framework for real-time computation over unbounded data.

  • Open source

Apache Kafka

Data Plane · Streaming & Stream Processing

Distributed event streaming platform for high-throughput data pipelines and streaming applications.

  • Open source

Azure Event Hubs

Data Plane · Streaming & Stream Processing

Managed event ingestion service on Azure, with a Kafka-compatible endpoint.

  • Cloud service

Redpanda

Data Plane · Streaming & Stream Processing

Kafka API-compatible streaming data platform written in C++.

  • Source available

Drafted with AI assistance and checked against the vendor’s own documentation.