streaming platform

Confluent

Confluent is the commercial platform around Apache Kafka. This profile covers what it does, how it is priced, and where it stops being the right choice.

What the product actually does

Confluent sells Kafka as a managed service plus the connective tissue around it: schema registry, connectors for hundreds of sources and sinks, stream processing with ksqlDB and Flink, and governance across all of it.

The managed service is the reason most teams buy. Kafka is a distributed log with a consensus protocol underneath, and operating it well means understanding partitions, replication, rebalancing and disk pressure. Confluent removes that from the team's plate.

How pricing works once you are inside

Confluent Cloud prices on throughput — a combination of ingress, egress and storage — measured in a unit it calls a CKU for dedicated clusters, or per gigabyte for the serverless tier. The serverless tier is cheap at low volume and becomes expensive at steady high volume, which is the opposite of most infrastructure pricing.

The metric that decides the bill in practice is egress. Data leaving the cluster for a sink is billed, and a pipeline that replicates into three destinations pays three times for the same bytes.

Where it breaks down

Cost at scale. A high-throughput pipeline on Confluent Cloud costs more than running Kafka yourself by a wide margin, and the break-even is somewhere around the point where a platform team is cheaper than the licence.

The other limit is that Kafka's model — a partitioned, ordered log — is not a universal fit. Workloads that need per-message routing or exactly-once semantics across heterogeneous systems end up building the guarantees themselves.

What switching costs

Moderate. Topics and schemas port to any Kafka; the connectors, the governance metadata and the stream processing jobs are Confluent-specific.

Sources

Confluent public documentation and pricing page, read October 2026.