What’s New in Confluent Cloud and WarpStream

Whether you’re building AI, scaling mission-critical Kafka workloads, or simplifying operations across your data estate, Confluent is innovating faster to help you do more on one fully managed data streaming platform.

Here are the latest Confluent Cloud and WarpStream updates, including expanded Kafka workload support and migration capabilities, new AI and Flink experiences, enterprise-ready security and USM enhancements, new fully managed connectors, and more.

Stream – Bring Any Kafka Workload to Confluent Cloud with Ease

• Expanded protocol compatibility in Freight, KStreams Debug View, and Queues for Kafka unify more workload patterns on one managed platform.
• Take on massive workloads up to 32 eCKUs on Enterprise and 252 CKUs on Dedicated. New metrics on throttle clients, connection and partition counts, along with follower fetching, Private Network Interface (PNI) networking, and max eCKU caps, give teams better cost and capacity controls as usage grows.
• Protect mission-critical workloads with release priority and cross-cloud Cluster Linking, then simplify the move to Confluent Cloud with KCP.

Confluent Intelligence and AI Tools – Making Agents Native to the Stream

• Confluent Intelligence: Real-time anomaly detection and forecasting with IBM Granite Time Series models and Google TimesFM model, now in Early Access. Get Started using the AI/ML Quickstart.
• AI Tools: Fully managed MCP server now supports actions such as restarting connectors and updating connector configuration. New Agent Skills include flink-udf, developing-kafka-java-client, msk-migration, and more. Confluent Copilot is now available through Early Access for building and operating Confluent Cloud.
• For more, Read the AI launch blog and Sign Up for Early Access.

Connect and Process – New Connectors and Making Flink Powerful and Accessible for All

• New fully managed connectors for Google Spanner CDC, Salesforce V2, and Google Cloud Firestore.
• Table API is now generally available, UDFs (GA) now support Java on GCP and Python on AWS, Process Table Functions (PTFs) are GA and bring Java support to all cloud providers, and new scaling controls provide a predictable performance floor for latency-sensitive workloads.
• Learn more in our Flink product blog.

Tableflow – Enterprise-Ready at Scale with Powerful New Capabilities

• External catalog integration and Iceberg REST catalog integration through Private Link, BYOK for Enterprise clusters, and observability features help teams monitor and maintain data lake pipelines at scale.
• Get Started with Tableflow.

Security and Unified Governance Made Simple – Extend Existing Controls to Confluent Cloud

• OAuth support for AWS IAM and Azure Managed Identities, IP filtering for Kafka, Secret Manager support for connectors, mTLS with certificate revocation, Global API Keys, TLS 1.3, and BYOK for Freight clusters.
• Private networking coverage now extends across Private Network Interface (PNI) egress for managed connectors and PrivateLink support for custom connectors. The Audit Log Monitoring UI and push-based integrations with Datadog and Splunk help security teams investigate activity in the tools they already use.
• Unified Stream Manager (USM) now supports Azure through Azure PrivateLink and adds native alerts for faster detection, triage, and resolution. Monitoring self-managed connectors on Confluent Cloud is in limited availability.

WarpStream – Improve Latency, Resilience, and Migration Flexibility

• WarpStream Lightning Topics acknowledge produce requests as soon as data is durably written to object storage, reducing median produce latency to 33 milliseconds when combined with S3 Express One Zone. Ripcord Mode keeps agents accepting produce requests even when the Control Plane is unreachable.
• Additional updates include: queryable Audit Logs, native Events for searching and visualizing operational activity, an MCP Server for natural-language cluster exploration, GCP Rapid Buckets for lower-latency object storage, and Orbit Auto Migration for gradual, per-topic Kafka cutovers without a coordinated maintenance window.
• Sign Up for WarpStream to try the newest features.

Data Streaming News

• Ravinder Panwar writes about 5 Event-Driven Anti-Patterns That Are Secretly Killing You Production (Apache Kafka) Systems.
• Sébastien Viale and Loïc Greffier of Michelin give of the details on Dead Letter Queue in Kafka Streams (KIP-1034).
• Last month, Aiven’s Olena Babenko argued that Apache Kafka Share Groups are NOT true queues (and why that’s not a a bad thing).
• Not a “data streaming” piece, but check out Markus Eisele’s discussion on the “right amount” of spec for agentic development.

In-Person Meetups

Here’s a list of meetups around the globe in the coming weeks:

• Aug 26th: Kuala Lumpur
• Aug 29th: Bangalore
• Sep 12th: Pune
• Sep 26th: Bangalore
• Sep 27th: Hyderabad
• Sep 29th: Singapore

See our Confluent Community Meetups page to keep up-to-date on opportunities to connect, learn, and network.

By the way…

We hope you enjoyed our curated assortment of resources!

If you’d like to view previous editions of the newsletter, visit our archive.

If you want to learn more about Kafka, Flink, or Iceberg, visit our developer site at Confluent Developer.