New architectural paradigm makes streams first-class lakehouse primitives; Ursa For Kafka delivers up to 95% cost reduction as the first proof point

StreamNative, founded by the creators of Apache Pulsar, today introduced Lakestream, a new architectural paradigm for lakehouse-native streaming, alongside the launch of Ursa For Kafka (UFK) — a native Apache Kafka service that puts the Lakestream vision into practice. After seven years building streaming infrastructure from the ground up, from Apache Pulsar's compute-storage separation to lakehouse-native storage, StreamNative is challenging that assumption from a place no connector-first vendor can.

The Streaming-Lakehouse Divide

Today's data architectures operate in a split world. Streaming data lives in Kafka or Pulsar. Analytics lives in the lakehouse like Databricks, Snowflake, or data lake platforms. Connecting the two requires layers of connectors, ETL pipelines, and materialization jobs that add cost, complexity, latency, and fragility.

The industry has tried to close this gap from both sides: Streaming vendors build connectors to the lakehouse, while lakehouse platforms add streaming features. But these approaches share a fundamental assumption: streaming and the lakehouse are separate systems that need to be bridged.

StreamNative is challenging that assumption.

Introducing Lakestream: Streams as Lakehouse Primitives

Lakestream is a new architectural paradigm that unifies data streaming and the lakehouse, not by building better bridges between them, but by making them one and the same. Just as the lakehouse paradigm proved that data warehouses and data lakes don't need to be separate, Lakestream proves that streaming and the lakehouse don't need to be separate either.

The core insight: traditional streaming systems push interoperability up to the protocol layer, creating data silos for each protocol. Lakestream pushes interoperability down to the storage and catalog layers - achieving unification through a shared lakehouse-native storage foundation and unified metadata catalog. Streams become first-class lakehouse primitives alongside tables. The protocol becomes a choice of interface, not a choice of data silo.

In practice, this means a Kafka topic and an Iceberg table can be the same object: no movement, no connectors, no waiting. The Lakestream architecture that makes this possible is built on three layers: cloud-native stream storage that writes directly to object storage in open formats (Iceberg, Delta Lake); a Lakestream Catalog that federates with Databricks Unity Catalog, Snowflake Horizon Catalog, and AWS S3 Tables; and stateless protocol servers that let Kafka, Pulsar, and other protocols all write to the same underlying storage.

Ursa For Kafka: Lakestream in Action

As the first major proof point of the Lakestream architecture, StreamNative is simultaneously launching Ursa For Kafka (UFK), a native Apache Kafka service entering Limited Public Preview. UFK is an Apache Kafka 4.2+ fork powered by Ursa, StreamNative's lakehouse-native stream storage engine. With UFK, every native Kafka topic is simultaneously a lakehouse table, queryable from Spark, Snowflake, and Databricks with zero code changes, no connectors, and no ETL.

Ursa's architecture eliminates cross-AZ replication (the single largest cost driver in cloud-deployed streaming) delivering up to 95% cost reduction validated at 5 GB/s sustained throughput. The architecture earned the Best Industry Paper award at VLDB 2025, selected over submissions from many industry leaders.

"We've spent years building toward this moment. Lakestream is the realization that streaming and the lakehouse aren't two systems that need to be connected, they're one system that was never designed to be unified – until now. Ursa For Kafka proves that even the world's most popular streaming protocol can become a native lakehouse citizen without sacrificing anything. The protocol stays the same. The data becomes lakehouse-native. The boundary simply disappears,” said Sijie Guo, CEO and Co-Founder, StreamNative

Key Capabilities of Ursa For Kafka

• Native Apache Kafka Protocol — Apache Kafka 4.2+ fork; every existing Kafka client, tool, and connector works with zero code changes
• Lakehouse-Native Storage — Every topic stored as Iceberg or Delta Lake tables on object storage; query streaming data from Spark, Snowflake, Databricks, and Trino
• Up to 95% Cost Reduction — Leaderless architecture eliminates cross-AZ replication, validated at 5 GB/s sustained throughput
• Zero-Connector Lakehouse Integration — No Kafka Connect, no materialization pipelines, no sink connectors; Kafka topics ARE lakehouse tables
• Catalog Integrations — Works with Databricks Unity Catalog, Snowflake Horizon Catalog, and AWS S3 Tables out of the box
• Mixed Storage Flexibility — Run cost-optimized (lakehouse-native) and latency-optimized (disk-based) topic profiles in the same cluster
• Available on AWS and GCP — With Azure expansion planned

"StreamNative and Starburst bring together real-time streaming and seamless Iceberg ingestion to simplify how data moves and becomes usable. Together, we create an open, AI-ready foundation. With Starburst, organizations can query that data directly at scale for real-time analytics and AI,” said Jitender Aswani - Senior Vice President, Engineering at Starburst.

"Real-time applications depend on fast, reliable streaming data. With StarTree’s native integration with StreamNative Kafka powered by the Ursa engine, organizations can transform event streams into real-time analytics. StarTree also enables customers to query Kafka data written as Iceberg tables by StreamNative, bringing real-time insights to open lakehouse environments,” said Chinmay Soman, VP Product at StarTree.

UFK demonstrates that any streaming protocol can become lakehouse-native through the Lakestream paradigm. The same architecture already powers Apache Pulsar workloads on StreamNative Cloud, and is designed to support any protocol that needs lakehouse-native streaming.

StreamNative will open source Ursa and key Lakestream components in the coming months, reflecting the company's belief that this architectural paradigm belongs to the community rather than any single vendor. Apache Pulsar continues to be fully supported on StreamNative Cloud for mission-critical messaging and streaming workloads.

Availability

Ursa For Kafka is entering Limited Public Preview today available on AWS and GCP through StreamNative Cloud. UFK works with existing Kafka clients version 0.9 and above — no code changes required.

Early enrollees who sign up by April 15, 2026 will receive $1,000 in credits to use exclusively for Kafka clusters on StreamNative Cloud. To learn more and sign up, visit StreamNative Cloud.

Additional Resources

Today’s announcement includes several launch partners that have validated technical integration. For more information on StreamNative’s launch partners, visit this link.

For more information on Lakestream, read StreamNative’s blog post.

For more information on UFK, read StreamNative’s blog post.

About StreamNative

StreamNative, founded by the creators of Apache Pulsar, is a leading provider of cloud-native streaming and messaging infrastructure. StreamNative Cloud enables enterprises to run Apache Pulsar and Apache Kafka workloads as fully managed services on AWS, GCP, and Azure. With the introduction of the Lakestream architecture and Ursa For Kafka, StreamNative is pioneering the unification of data streaming and the lakehouse — making streams first-class lakehouse primitives alongside tables. StreamNative is backed by investors including Prosperity7, Tencent, and others, and serves customers across financial services, technology, telecommunications, and more.

View source version on businesswire.com: https://www.businesswire.com/news/home/20260407143701/en/