ClickHouse Accelerates: AI, Streaming, and Cost-Performance Unveiled

ClickHouse's Strategic Leaps: AI, Compute, and Data Integration

The ClickHouse September 2026 newsletter showcases a vibrant development ecosystem with a strong focus on expanding capabilities into AI and improving operational efficiency. The introduction of On-Demand Compute, coupled with a new cost-based optimizer and distributed query execution framework, is a major step towards making ClickHouse a more versatile platform for intensive, ad-hoc, and data lake workloads. This directly addresses a key challenge in data warehousing: balancing performance with cost for unpredictable or high-demand tasks. The integration of AI functions is another significant indicator of ClickHouse's commitment to modern data analytics needs, enabling in-database machine learning operations. Furthermore, the continued emphasis on PostgreSQL integration, particularly through WalShadow and chdb, solidifies ClickHouse's position as a strong contender in heterogeneous data environments and for real-time replication scenarios.

However, the 'private preview' status for several key features like On-Demand Compute and AI functions means widespread adoption and real-world impact are still some way off. While the newsletter details the 'what' and 'why,' more depth on the 'how' for these advanced features in terms of configuration, resource management, and potential pitfalls would be beneficial for early adopters. The CostBench results are compelling, but a deeper dive into the methodology and the specific architectural choices that lead to ClickHouse's performance-per-dollar advantage would further bolster its credibility. The sheer volume of content also suggests a rapid development cycle, which, while exciting, can sometimes lead to concerns about stability and long-term support for individual features, especially with 57 breaking changes noted in the 26.8 LTS release since 26.3.

Key Points

• ClickHouse 26.8 release introduces custom HTTP handlers, pipelined SQL ('|>'), and Japanese/Chinese tokenizers for text-index.
• On-Demand Compute is entering private preview, aiming to offload intensive workloads with a new cost-based optimizer and distributed query execution framework.
• New AI functions are available, enabling in-database AI model integration for tasks like text generation and embeddings.
• Significant advancements in PostgreSQL integration, including WalShadow for direct WAL replication to ClickHouse and the chdb Postgres extension.
• CostBench results show ClickHouse Cloud's strong performance-per-dollar under continuous load compared to competitors.
• Community contributions highlighted, including Rory Shanks' improvements to replication and JSON subcolumn reads.
• Upcoming events and training sessions cover a wide range of topics from query optimization to AI agents and PostgreSQL integration.

---

📖 Source: September 2026 newsletter