ClickHouse Advances Observability, AI Features and Global Expansion in Busy Week

ClickHouse entered the week spotlighting its role in next-generation observability as digital bank Qonto adopted its cloud platform to shift from traditional logs, metrics, and traces to “wide events” in a single system. The deployment reportedly expanded query windows from hours to weeks and enabled higher-cardinality data retention, while an AI-driven incident companion built on the ClickHouse MCP server lets staff investigate production issues using natural language.

In parallel, ClickHouse highlighted ecosystem performance gains through a Streamfold Rotel case study, where an optimized OpenTelemetry Collector achieved about 3.7 million trace spans per second on standard hardware versus roughly 1.1 million for a standard setup. The improvements were attributed to fixes for glibc arena lock contention, RowBinary column formats, Tokio-based task parallelization, and LZ4 compression, underlining ClickHouse’s suitability for petabyte-scale telemetry workloads.

Product development remained a major focus with the v26.4 release of its cloud-native analytics database, presented by founder Alexey Milovidov from Japan in a live community call. New capabilities include automatic spilling to disk for JOIN operations to prevent out-of-memory errors, Arrow Flight SQL support for higher-performance ingestion, and native AI functions, aiming to strengthen real-time analytics and AI use cases.

The company also emphasized enhancements in version 26.3, where asynchronous inserts became the default, buffering data in memory and flushing in larger batches to reduce “Too many parts” errors at high ingest rates. Configurable acknowledgment modes and flush triggers are designed to give users flexibility between durability and throughput, while internal benchmarks cited up to 5x faster DISTINCT operations on low-cardinality columns and 5x faster Parquet reads on repeat queries.

Tooling investments were showcased through clickhousectl, an internal command-line utility that simplifies installing and switching between versions, running multiple local servers, and moving schemas and data without containers. These capabilities are intended to lower adoption friction for developers and data teams, potentially speeding experimentation and proof-of-concept cycles around the platform.

Strategically, ClickHouse announced the official launch of ClickHouse Japan with an event in Tokyo featuring founders, global leadership, and the local team, signaling a push into one of the world’s most important technology markets. Taken together, the week’s updates on observability use cases, performance improvements, AI features, and geographic expansion point to a company intensifying its product and go-to-market efforts in high-performance analytics and data infrastructure.