Confluent rolled out a broad set of updates across Confluent Cloud and WarpStream covering Kafka protocol support, scaling limits, networking, migration tooling, connectors, Flink, Tableflow, AI capabilities, security, and observability. Highlights include transactions and idempotent produce on Freight clusters, Python share consumers for Queues for Kafka, higher eCKU/CKU scaling limits, automatic Private Network Interface setup, the KCP migration tool now supporting any Apache Kafka source plus client migration, new Debezium and Salesforce/Firestore connectors with discounted Connect Advantage pricing, GA of Flink Table API for Java on GCP and Python on AWS, Unified Stream Manager expansions (Kafka Streams UI, Azure support, native alerts, connector monitoring), Tableflow private connectivity and BYOK, new OAuth/mTLS/IP filtering security features, and WarpStream additions like Lightning Topics (33ms median produce latency), Ripcord Mode for control-plane outages, Audit Logs, Events, an MCP server for AI assistants, GCP Rapid Buckets, and Orbit Auto Migration for zero-downtime cutovers.
Table of contents
Stream: Bring Any (Kafka) Workload, at Any Scale to Confluent Cloud with Confidence and EaseConnect: Connect with 100+ Systems and Save More with Predictable UsageProcess: Powerful and Accessible Flink for Both Developers and Data PractitionersGovern: Unified, Schematized Governance for Hybrid Confluent EnvironmentsTableflow: Now Enterprise-Ready at Scale with Powerful New CapabilitiesAI: Making Agents Native to the StreamSecurity: Extend Your Existing Security Controls to Confluent CloudWarpStream: BYOC Diskless Kafka Keeps Getting BetterLightning Topics and Ripcord ModeAudit LogsEventsMCP ServerGCP Rapid BucketsOrbit Auto MigrationStart Building with New Confluent Cloud and WarpStream FeaturesQuestions this post answers
How much does WarpStream's Lightning Topics feature reduce produce latency?
Lightning Topics cuts median produce latency to 33ms, a 70% reduction, by acknowledging produce requests as soon as data is durably written to object storage and deferring offset sequencing to an asynchronous background step. Combined with S3 Express One Zone, this makes WarpStream viable for latency-sensitive workloads as a diskless Kafka-compatible product. Teams weighing diskless Kafka options for low-latency workloads can track releases like this on daily.dev.
What happens to WarpStream writes if the control plane becomes unreachable?
WarpStream's Ripcord Mode lets Agents keep accepting produce requests even when the Control Plane is completely unreachable, journaling records to object storage and committing them asynchronously once connectivity is restored. This removes the Control Plane as a hard dependency for writes, keeping applications that cannot backpressure or drop data running through networking issues or Control Plane incidents. Engineers designing for streaming resilience can follow updates like this via daily.dev.
What is the maximum cluster size Confluent Cloud supports for Dedicated clusters?
Confluent Cloud now scales Dedicated clusters up to 252 CKUs on AWS (upon request), or 152 CKUs on Google Cloud and 100 CKUs on Azure, delivering up to 15 GB/s ingress and 45 GB/s egress. Enterprise clusters scale up to 32 eCKUs, supporting the most demanding Kafka workloads without redesigning around infrastructure limits. Teams planning capacity for large Kafka deployments can follow scaling limit changes on daily.dev.