- March 1, 2026: Stricter limits for requests on Enterprise and Freight clusters.
- June 30, 2026: Stricter limits for total client connections and connection attempts on Enterprise and Freight clusters.
- June 30, 2026: Stricter limits for requests, total client connections, and connection attempts on Basic and Standard clusters.
Practically, teams that rely on heavy request throughput or many concurrent client connections may need to scale up to more eCKUs to avoid throttling. Cluster max eCKU limits are 50 for Basic, 10 for Standard, 32 for Enterprise, 152 for Freight. The change is meant to align cloud-native Kafka with on-prem limits but creates migration friction for some workloads (Confluent docs, August 2026).
Hidden cost dimensions
The headline eCKU rate is the visible cost; the hidden line is per-dimension usage. Confluent Cloud bills across multiple dimensions: eCKUs (compute capacity), ingress (data written), egress (data read), storage (data held), and per-feature meters for connectors, ksqlDB, Schema Registry, and Stream Governance.
Connector tasks bill per task-hour plus throughput ($/GB). For fully managed connectors like Kafka Connect JDBC or Debezium CDC, this adds $0.10-$0.50/task-hour. Schema Registry is free for the first 1,000 schemas, then billed per schema-month. ksqlDB streaming units (CSUs) bill per CSU-hour. Stream Governance bills per environment-hour (Confluent docs, August 2026).
Two cost-control levers: Volume-Based Discounts are negotiated upfront based on planned annual consumption; Automatic Tiered Pricing applies deeper discounts as usage grows without renegotiation. Most teams benefit more from Automatic Tiered Pricing because it triggers automatically. Enterprise contract negotiations typically unlock 15%-30% off the headline eCKU rate.
When to choose which Confluent tier
Choose Basic for development, testing, and basic use cases with 99.5% SLA tolerance. Choose Standard for production workloads using public networking with 99.9% (single eCKU) or 99.99% SLA (2+ eCKUs). Choose Enterprise for production workloads requiring private networking (VPC peering, PrivateLink), GBps+ autoscaling, and 32 eCKU max. Choose Freight for cost-optimized high-scale logging, observability, batch analytics, and any use case that tolerates relaxed latency in exchange for lower cost (Confluent pricing page, August 2026).
What enterprise buyers should do next
- Model your traffic pattern against eCKU limits. Standard maxes at 10 eCKUs (250 MBps ingress); Enterprise at 32 eCKUs (1,920 MBps); Freight at 152 eCKUs (9,120 MBps). Pick the tier that matches peak traffic.
- Audit free partition counts. Basic includes 10 free partitions, Standard 500, Enterprise unlimited. If your topic count exceeds free partitions, plan for overage.
- Negotiate annual commit for 15%-30% off. Confluent's eCKU rate is negotiable for annual commits above $100,000.
- Calculate total cost including connectors, ksqlDB, and storage. Per-dimension charges for connectors and storage can double the headline eCKU bill.
What to watch next
Three near-term datapoints. First, the June 30, 2026 stricter limits enforcement — Basic and Standard clusters will hit request/connection limits that previously were lax. Second, Apache Kafka 4.0 GA timing — Confluent Cloud typically rolls out new Kafka features within 60 days of GA, and pricing for new features like KIP-848 (next-gen consumer group protocol) may shift. Third, Confluent's Tableflow pricing — Tableflow bills per topic-hour and per GB processed, which can add to total Kafka TCO for Iceberg/Delta Lake integrations.