Ververica Platform 3.0: The Turning Point for Unified Streaming Data

The world of data has been changing rapidly. In 2025, businesses need to make decisions in real time, deliver personalized customer experiences instantly, and enable applications that learn and adapt on the fly. Yet most data platforms remain fragmented, forcing teams to stitch together ingestion, processing, storage, and analytics across a patchwork of systems.

On top of that, companies are often forced to choose between batch or streaming approaches. Batch systems provide reliability for historical analysis, while streaming delivers low-latency insights, but each lives in its own silo, with different tooling, infrastructure, and results. This fragmentation leads to complexity, higher costs, and slower innovation.

With the launch of Ververica Platform 3.0, we’re changing that. This release represents a turning point: Ververica has evolved into a complete Unified Streaming Data Platform. For the first time, you can connect, process, store, analyze, and govern your data end-to-end, all in a single platform designed for the speed and scale of modern data applications without compromise, and without choosing between batch and streaming.

A Platform Built for the Future of Data in the Age of AI

From its origins as the commercial home of Apache Flink®, Ververica stands at the forefront of stream processing. With Ververica Platform 3.0, we are extending that heritage into something broader, a foundation that combines real-time stream processing, streaming storage, and advanced analytics with the governance and security enterprises demand.

This new release introduces three major innovations that bring the vision to life:

VERA-X: Redefining Stream Processing

At the heart of the release is VERA-X, the new stream processing engine. VERA-X is the first native vectorized execution engine for Apache Flink, a leap forward in performance and efficiency. By optimizing how Flink executes queries at a low level, it delivers 5–10× higher throughput and single-digit millisecond latency on streaming benchmarks, while cutting infrastructure costs by up to 50%. 

VERA-X is also not just for streaming. Batch queries run 2–3× faster than Flink’s Java runtime and up to 3.1× faster than Apache Spark™ 3.4.3, all with zero changes to existing Flink applications.

For developers and operators, this means you can continue to use familiar Flink APIs and SQL while benefiting from a next-generation engine that’s built for the workloads of tomorrow.

Apache Fluss: Completing the Streamhouse Vision

The Streamhouse unifies real-time streaming with the analytical power of the lakehouse. Flink provides best-in-class processing, but what is  missing is a storage layer purpose-built for streaming. Apache Fluss fills this gap.

Fluss treats streams as queryable tables, continuously updated and instantly accessible, so live data can be processed and analyzed the moment it arrives. This eliminates fragile pipelines and reduces the need for duplicate systems.

In Ververica Platform 3.0, Fluss works side by side with open table formats such as Apache Paimon™ and Apache Iceberg™. Fluss keeps the fresh, real-time data always available, while Paimon or Iceberg provide the historical persistence needed for large-scale analytics and governance. Together, they bridge the worlds of operational and analytical data.

This combination brings clear benefits:

• Unified view of live and historical data without stitching multiple systems together.
• Efficiency through less duplication and reduced infrastructure overhead.
• AI-readiness with vector search, real-time feature stores, and retrieval-augmented generation (RAG) directly on streaming data.

With VERA as the engine, Fluss as the live data store, and Paimon or Iceberg for historical data, Ververica Platform 3.0 delivers the complete Streamhouse vision — one cohesive platform for data in motion and at rest.

Real-Time AI With Rag And LLM Support

Another exciting development in 3.0 is the arrival of real-time AI capabilities directly within the platform. With new retrieval-augmented generation (RAG) and large language model (LLM) support in Flink SQL, you can now bring the power of AI into your streaming pipelines.

This means enriching live data with context-aware insights, powering conversational interfaces that respond to the most recent events, or enabling AI-driven decision-making at the moment data arrives. Real-time AI is no longer a future vision, it’s something you can build today on Ververica Platform 3.0.

No More Choosing Between Batch and Streaming

For years, organizations had to choose: build for batch or build for streaming. Batch systems offer reliability for historical data analysis, while streaming systems provide low-latency insights. But each lived in its own world, with different tooling, data models, and infrastructure. With Ververica, that complexity is removed, as both batch and streaming pipelines are unified into a singular solution.

At the heart of everything we do at Ververica is a streaming-first design philosophy. Whether your workload is continuous streaming or large-scale batch analytics, both are treated as first-class citizens to Apache Flink. This ensures consistency across workloads, eliminating the old divide between “real-time” and “historical” systems.

Image 1: Unified Streaming-First Data Platform for AI

With Ververica Platform 3.0, any trade-offs disappear. By unifying batch and streaming, you gain:

• One platform, one model: Whether data is live, or historical, developers now work with a single API and SQL dialect.
• Consistent results: between batch-processed reports and real-time dashboards disappear, as both are powered by the same engine.
• Simplified operations: Reduce complexity and operational overhead with a single unified runtime and storage layer.
• Faster innovation: Teams can combine real-time and historical context seamlessly, powering richer analytics and more intelligent applications.
• Cost efficiency: Unified infrastructure means less duplication and fewer moving parts to manage.

This convergence of batch and streaming is what makes Ververica Platform 3.0 truly powerful - the best of both worlds, working together.

The Value of a Unified Approach

Ververica Platform 3.0 is more than a product release. It represents a shift in how organizations can think about their data. Instead of managing separate systems for ingestion, processing, storage, analytics, and AI, you now have a single platform that does it all.

For enterprises, this means:

• Simplicity: Fewer moving parts and tighter integration across the data stack.
• Performance: A next-gen engine that scales with your needs.
• Flexibility: Support for both operational and analytical workloads in one place.
• Innovation: Built-in AI capabilities that allow you to experiment and deploy faster.
• Governance and trust: Enterprise-grade security and management, ensuring your data is protected.
• Unified data experience: No more silos between real-time and batch processing. Everything comes together in a single platform.

A Turning Point for Ververica

Ververica Platform 3.0 represents a new era for both us and our users. From its roots in Apache Flink to the innovations of VERA-X and Fluss, we’ve always believed in the power of data in motion. Now, with the unification of batch and streaming, and the arrival of real-time AI, that belief becomes reality.

This is the most complete expression yet of our vision: a streaming-first, unified data platform that powers the applications of tomorrow.

And this is only the beginning. Ververica Platform 3.0 is not just a turning point for our product; it is the foundation of real-time AI. Because AI that truly matters is AI that acts on the freshest data, adapts instantly, and learns continuously. That requires more than static snapshots of the past; it requires a platform where streams and historical context come together seamlessly, where data is always alive, and where intelligence is built into the flow.

That is what Ververica Platform 3.0 delivers. The future of real-time AI starts here.