Data warehouses
Each provider is profiled from its own pricing page, with the source recorded against every figure. One charges for the bytes a query reads, one for the hours a cluster runs, one for the hours a service is awake and one for the organisation — so the column that matters is what the price counts, and beside it what storage costs on top.
Data warehouses
| Provider | Entry paid plan | Storage price | What is metered |
|---|---|---|---|
| Amazon Redshift | $0.543 an hour for a provisioned cluster, or $1.50 an hour for serverless — about $391 and $1,080 a month if left running. Serverless shuts down when idle and bills per second with a 60-second minimum, so the higher hourly rate can be the cheaper one. | Redshift Managed Storage grows independently of compute, so a cluster can be sized for query performance without over-provisioning disk. The page states the separation as the design point rather than quoting one rate for it. | Two deployment options priced on time rather than data. Provisioned bills per node-hour for a cluster you size yourself; Serverless bills Redshift Processing Unit hours per second, with a 60-second minimum and no charge for start-up time, shutting down when inactive. Managed storage is billed independently of compute in both. |
| ClickHouse Cloud | From $66.52 a month for a Basic service active 6 hours a day — one replica with 8 GiB of RAM and 2 vCPUs, 500 GB of compressed data and 500 GB of backup. The same service active 24 hours a day is $186.27, so the entry price is a statement about usage as much as about capacity. | Not documented | Compute, storage and data transfer itemised separately, with compute varying by how many hours a day the service is active — $39.91, $79.83 or $159.66 in the published example for 6, 12 or 24 hours, against a flat $25.30 of storage. Scale, the always-on auto-scaling tier, starts at $499.38 a month. |
| Google BigQuery | Nothing until the second tebibyte. Queries are $6.25 per tebibyte scanned above the free tier, with no cluster, node or plan fee underneath — the only member here where a small workload genuinely costs zero. | $0.000031507 per gibibyte-hour for active logical storage and $0.000021918 for long-term, with the first 10 GiB free each month. Quoted hourly to nine decimal places; the page does not convert it to a monthly figure and neither does this record. | Two compute models. On-demand bills the bytes each query processes, at $6.25 per tebibyte. Capacity bills slot-hours — reserved virtual CPUs — regardless of what the queries read. The first model punishes unselective queries and the second punishes idle capacity, and the choice is made per account rather than per query. |
| MotherDuck | $250 per organisation per month, plus usage. Priced on the organisation rather than on data or time — the only member here where adding a query, a gigabyte or an hour does not change the floor. | Not documented | Per organisation per month plus usage, with seats rather than scans as the limiting dimension: up to 10 internal active users on Business against 3 on Lite, and unlimited service accounts. Five instance types and read-scaling replicas sit above that. |
Which to consider
Constant, predictable query load already inside AWS. At $0.543 an hour a provisioned cluster costs the same whether it serves one query or a million, so heavy continuous use is where the hourly model pays.
Knowing what a part-time warehouse costs. The published example is explicit — $66.52 a month for a Basic service awake six hours a day against $186.27 awake around the clock — with compute, storage and egress itemised separately.
Queries that are occasional or selective. Nothing is provisioned, the first tebibyte scanned each month is free, and a workload that runs twice a week costs nothing between runs — the only member here where a small warehouse genuinely bills zero.
Analytics that fit on one machine, priced per organisation rather than per query: $250 a month flat plus usage, so adding a query, a gigabyte or an hour does not move the floor. A 99.9% SLA comes with it.