Confluent is the commercial data streaming platform built around Apache Kafka by its original creators, sold as self-managed Confluent Platform and as the Confluent Cloud service, and acquired by IBM in March 2026.
Two shapes: Confluent Platform, a self-managed distribution you run and license, and Confluent Cloud, a managed service where eCKUs autoscale and Dedicated CKUs pin capacity. Managed connectors, tiered storage, and Stream Governance are platform features; Flink SQL and ksqlDB are billed separately from the cluster.
How Confluent answers the questions Event Streaming turns on.
| How it works | |
| Protocol | Kafka |
| Consumption model | Replayable log with offsets; Cloud added queues in 2026 |
| Storage | Kafka's, with tiered storage in Cloud |
| Delivery guarantee | At-least-once by default; Exactly-once with transactions |
| Running it | |
| What you run | Confluent Platform yourself, or nothing on Confluent Cloud |
| Scaling | eCKUs autoscale; Dedicated CKUs are fixed capacity you set |
| Multi-region | Cluster Linking between clusters, billed as its own dimension |
| Connections | |
| Connectors | Managed connectors on Cloud; Kafka Connect on Platform |
| Transformations | Flink SQL and ksqlDB, billed separately from the cluster |
| Cost | |
| Billing unit | eCKU or CKU hours, plus ingress, egress and 3× replicated storage |
vs Confluent: Managed · Serverless · Subscription · Operational complexity: Low
vs Confluent: Open source (permissive) · Self-hosted · Managed · Hybrid · Free · Operational complexity: High · Java, Scala
vs Confluent: Open source (permissive) · Self-hosted · Managed · Hybrid · Free · Operational complexity: High · Java