ClickHouse says its cloud service delivered 412× to 1,996× better end-to-end performance per dollar than four rival analytics services in its latest CostBench comparison. The eye-catching result comes from a test that measured both the cost of preparing incoming data and the speed and cost of queries, rather than timing queries in isolation.
ClickHouse Watch analysis
What happened
The company streamed 113.2 billion stock-market quotes into ClickHouse Cloud, Snowflake, Databricks, BigQuery and Redshift Serverless while analytical queries ran. ClickHouse reports the lowest preparation cost, query cost and accumulated query runtime across the comparison. Its score showed a 412× advantage over Snowflake, 752× over Databricks, 438× over BigQuery Capacity and 1,996× over Redshift Serverless.
ClickHouse says preparing the data cost $28.69 on its service, against about $79 for Snowflake, $243 for BigQuery Capacity, $522 for Redshift Serverless and $696 for Databricks. The company also says ClickHouse was the only tested system whose daily summaries stayed query-ready alongside the raw data. Its benchmark code, configurations and reproduction instructions are available, and the latest post explains how ClickHouse Cloud handles preparation, queries and scaling.
Key findings
- The score covers the whole analytical path
CostBench combines preparation and query-serving costs with accumulated query runtime, so faster answers have to justify their cost. - ClickHouse reports the lowest data-preparation bill
Preparing 113.2 billion rows cost $28.69 in its test, versus about $79 to $696 on the four alternatives. - Fresh summaries were part of the comparison
ClickHouse says its daily summaries stayed in sync with raw data, while the other tested systems refreshed summaries asynchronously.
Why it matters
For teams running real-time analytics, data preparation is not backstage housekeeping: its cost and freshness affect what queries can deliver. A benchmark that includes those steps is more useful than a neat query-speed chart that leaves the bill for ingestion and maintenance offstage.
The headline ratios are still ClickHouse’s results from its own benchmark, not a universal promise that every workload will be cheaper or faster. The company provides materials for reproducing the test, giving operators a way to check whether the claimed advantage survives contact with their data and service configuration.
Our read
This is a meaningful comparison because it measures cost and speed together, and makes the methodology available to inspect. The multiples deserve attention, not automatic adoption: teams should reproduce the workload with their own data and include the operational costs their own setup requires. Benchmarks are most persuasive when the reader can bring a calculator and a slightly sceptical eyebrow.
What to watch
- Whether independent reproductions find similar cost and freshness differences.
- How results change with other data shapes, query mixes and service configurations.
- Whether ClickHouse’s published benchmark materials make the comparison straightforward to reproduce.
Discussion spark: Should database benchmarks judge systems by query speed alone, or is a whole-pipeline measure of cost and freshness the fairer test?
Sources and evidence
- Source update (28 October 2026, 06:10 UTC)
not affiliated with or endorsed by ClickHouse