ClickHouse says its cloud service delivered 438 times better end-to-end performance per dollar than BigQuery on a continuous-ingestion benchmark, rising to 512 times under BigQuery’s on-demand pricing. The comparison ran queries while streaming 113.2 billion stock-market quotes, but the headline score combines costs with query runtime, so it is not a simple price-per-query result.
ClickHouse Watch analysis
What happened
ClickHouse’s benchmark report compares ClickHouse Cloud with BigQuery as both ingest data and answer queries. The workload streamed quotes at a target of one million rows per second, with four aggregate queries running every ten minutes and two raw-data drill-downs every hour. Result caching was disabled.
ClickHouse reports a combined performance-per-dollar score 438 times better than BigQuery using Capacity pricing, or 512 times better using On-demand pricing. Its formula combines preparation cost and normalised query cost with accumulated query runtime. The report’s detailed figures show ClickHouse at 58.29 seconds of accumulated query runtime, compared with 49.36 minutes for BigQuery. These are ClickHouse’s measurements and scoring choices, not an independent comparison.
Key findings
- The score includes speed as well as cost
ClickHouse’s formula multiplies preparation and query costs by accumulated query runtime, so the headline ratio is not a standalone bill comparison. - Both systems ran under a continuous workload
The test streamed 113.2 billion quotes while aggregate and drill-down queries repeatedly included newly arrived data. - The pricing scenarios are alternatives
Capacity and On-demand apply different prices to BigQuery resources used in the same measured workload; Capacity was not a separate run. - The comparison has explicit exclusions
The report excludes items including storage, discounts, idle or minimum-capacity charges, producer infrastructure and network costs.
Why it matters
Real-time analytics is not just a race to ingest data. Systems must prepare it for queries while keeping answers current, and the costs of preparation, querying and runtime can pull in different directions. ClickHouse’s report attempts to measure that whole path rather than presenting a single query in isolation.
The scale and detail make the result worth examining, but the headline needs its footnotes close by. BigQuery used dynamically assigned serverless slots, while ClickHouse used a specified Cloud configuration; the report says this is not a comparison of two fixed 16-CPU allocations. ClickHouse also acknowledges a small difference between the providers’ successful-write counts. The result describes this workload and method, not every analytics job or a typical provider invoice.
Our read
This is a useful, unusually detailed benchmark, and the runtime gap is striking. But 438× is a composite score, not a magic multiplier to apply to your own bill. Anyone choosing a platform should inspect the workload, pricing assumptions and exclusions, then test their own queries and data. Benchmarks are most useful when they arrive with enough working to check the arithmetic.
What to watch
- Whether independent tests reproduce the performance and cost differences on comparable real-time workloads.
- How the result changes with different query mixes, data shapes and ingestion rates.
- Whether ClickHouse publishes further benchmark runs that make the exclusions and provider costs easier to compare.
Discussion spark: When choosing a real-time analytics platform, should buyers give more weight to a broad performance-per-dollar score or to a benchmark built around their own workload?
Sources and evidence
- Source update (8 October 2026, 10:49 UTC)
not affiliated with or endorsed by ClickHouse