ClickHouse Pricing 2026: The Real Cost of Running It 24/7

ClickHouse Cloud’s Basic tier costs $66.52 a month at six active hours a day, and $186.27 a month at 24/7. Same tier, same single 8 GiB replica, same data. Uptime is the only variable that moved.

Most ClickHouse pricing guides stop at the rate card, $0.2181 per compute unit-hour and $25.30 per compressed terabyte, and leave the arithmetic to you. This one does the arithmetic: real monthly totals for all three ClickHouse Cloud tiers, what self-hosting costs once the omitted lines are counted, and which option is cheapest for a given workload.

Selfhost.dev runs managed ClickHouse on dedicated instances, so these are bills we model regularly. Every ClickHouse Cloud figure below was verified against clickhouse.com on August 8, 2026, and where our reading differs from figures in wide circulation, the working is shown.

If you are sizing a ClickHouse deployment, you will finish this able to name your deployment model and the two variables that decide it.

Key takeaways

• Uptime is the price. ClickHouse Cloud Basic runs $66.52 a month at six hours a day and $186.27 at 24/7. Compute is the only line that moves, and it moves 4x.
• The tiers gate scaling, not capacity. Basic cannot scale at all, manually or automatically. The next tier that can, Scale, starts at $499.38.
• Compute is $0.22 to $0.39 per unit-hour, not the $0.75 widely quoted. One compute unit is 8 GiB RAM and 2 vCPU.
• Storage is $25.30 per compressed TB per month and identical on every tier, with backups billed separately at the same rate.
• Self-hosted estimates in circulation differ by roughly 16x, from about $150 to over $2,400 a month, because most fold an engineering-time assumption into the infrastructure figure.
• An always-on dedicated instance runs about $64 a month in Mumbai, against $175 on ClickHouse Cloud Basic for one replica at the same 8 GiB, 2 vCPU specification with 100 GB of data.

How much does ClickHouse cost?

ClickHouse is free. The ClickHouse database is released under the Apache 2.0 licence, with no licence fee and no revenue share. What costs money is running ClickHouse: managed ClickHouse Cloud starts at $66.52 a month, third-party managed providers price per vCPU or as flat monthly plans, and self-hosting converts the bill into cloud infrastructure plus operational time.

ClickHouse Cloud publishes three tiers. The important thing to understand early is that the tier controls what your service is permitted to do, not how much data it can hold.

Basic is where most evaluations start and where most surprises live. Basic services cannot scale, manually or automatically, and memory allocation is not adjustable on the Basic tier. ClickHouse prices the Basic example at 500 GB of compressed data plus 500 GB of backup, and a Basic service is single-zone, so there is no failover.

Enterprise pricing is quoted rather than published, so treat $2,669.40 as the floor of the published examples rather than a list price.

The two alternatives to ClickHouse Cloud, self-hosting and third-party managed providers, follow different cost models entirely and are priced out further down.

One thing worth settling before you budget: ClickHouse Cloud has no free tier. New organisations get a 30-day trial with $300 in credits, and those credits expire at the end of the 30 days even if a payment method has been added. At least one guide currently ranking for this term describes a permanent free tier with 1 GB of storage. That is the trial, not a tier.

Three published prices still will not tell you what your own workload costs, because the total moves with uptime, compressed data volume and egress. You can price a configuration without creating an account using the Selfhost.dev pricing calculator, no signup needed.

How ClickHouse Cloud pricing works: the four meters

ClickHouse Cloud bills on four independent meters: compute, compressed storage, data transfer and ClickPipes ingestion. Compute is the one that decides most bills. ClickHouse Cloud meters compute per minute in units of 8 GiB RAM and 2 vCPU, and charges only while a service is actually running, which is why two teams on the same tier with the same data can pay very different amounts.

Two details in that table catch people out.

Storage is priced identically on every tier, so moving from Basic to Enterprise does not change what a terabyte costs. It changes what your service is allowed to do. And backups are billed separately at the same $25.30 rate, with every service defaulting to one backup retained for a day. That default quietly doubles the storage line, and each additional retained copy adds another full copy of your compressed footprint.

Compression is the other variable worth planning around. ClickHouse describes roughly 10x compression as typical for analytical data, but the real ratio depends on cardinality and codec choice, so the storage figure stays an estimate until you have actually ingested your dataset.

Four independent meters, each moved by a different thing, is what makes a consumption bill hard to forecast in advance. If you are trying to put a number on that uncertainty before you commit, it helps to have a way to measure billing predictability.

Of the four, uptime is the single strongest predictor of what you will pay. Two details in that table catch people out.

What ClickHouse Cloud costs at 6, 12 and 24 hours a day

The same ClickHouse Cloud Basic service costs $66.52, $106.44, or $186.27 a month depending only on how many hours a day it runs. Compute is the only line that moves, and it moves 4x across that range while storage and transfer stay flat.

Here is ClickHouse’s own breakdown for a Basic service with one 8 GiB replica, 500 GB of compressed data plus 500 GB of backup, 10 GB of public egress, and 5 GB of cross-region transfer in AWS us-east-1.

The $66.52 headline assumes the database is switched off for 18 hours out of every 24. Basic services pause after a period of inactivity by default, and that default is what makes the low end reachable. It is also what makes it awkward for anything user-facing, because a paused service has to come back before it can answer. The cheapest configuration and an always-available one are not the same configuration.

For a production analytics service that has to respond whenever someone opens a dashboard, $186.27 is the real number.

Why Basic cannot scale, and what Scale costs instead

Basic cannot scale at all. A Basic service is pinned to one replica at 8 GiB RAM and 2 vCPU, memory allocation is not adjustable, and deployment is single-zone with no failover. ClickHouse is direct about this in the console: to add RAM, add CPU, or add replicas for redundancy, the instruction is to switch to Scale or Enterprise.

Going from an always-on Basic service to ClickHouse’s published Scale entry example is a 2.7x jump, from $186.27 to $499.38, and the trigger is not data growth. It is needing a second replica, an extra availability zone, or automatic scaling during a spike. The $499.38 figure buys two replicas, which is the point: redundancy is the reason to move.

Scale’s published examples show where it goes from there:

Scale also raises the included transfer allowances, from 10 GB to 100 GB of public egress and from 5 GB to 10 GB cross-region.

The cliff, not the entry price, is the budgeting risk. A team that sizes a project against $66.52 and then discovers it needs failover is looking at a 7.5x increase, and no amount of query tuning closes that gap because the cost is structural.

There is a second route onto that cliff. The tier belongs to your ClickHouse Cloud organisation rather than to a single service, so buying their managed Postgres above 8 GB of RAM moves the whole organisation to Scale and takes your analytics service with it. The trigger sits lower than most teams expect, which we map out in the ClickHouse Managed Postgres rate card.

Our recommendation for always-on analytics: If your ClickHouse workload runs 24/7, consumption pricing becomes most expensive precisely when you need continuous availability. Enabling the features expected in production, such as multi-zone availability and additional replicas, also means moving to a tier that costs about 2.7x more than the entry tier.

A dedicated instance follows a different model: you pay a fixed hourly rate for dedicated hardware, so your monthly infrastructure cost stays predictable instead of increasing with uptime or forcing you into a higher pricing tier.

Deploy a dedicated ClickHouse instance if your analytics platform is always on, your traffic is predictable, and you want consistent performance with stable, transparent pricing.

What self-hosting ClickHouse actually costs

Self-hosting ClickHouse costs nothing in licence fees. ClickHouse is released under Apache 2.0, free for commercial use with no revenue share. The infrastructure for a single always-on node with 100 GB of storage runs about $42 a month at AWS list prices in Mumbai. Estimates above $2,000 are describing something else.

Here is what the infrastructure actually costs, using AWS on-demand list prices in ap-south-1:

Those figures cover compute and storage only. A public IPv4 address, egress, backup storage, and monitoring sit on top, and the same node in us-east-1 comes to $57.06 because AWS compute is roughly a third more expensive there.

The line that produces the $2,000 estimates is not a price. Most published self-hosted figures include an allocation for engineering time, usually 10 to 20 percent of an engineer, which lands between $2,000 and $4,000 a month depending on the salary assumed. A guide quoting $2,400 for a small cluster is quoting a few hundred dollars of infrastructure and roughly $2,000 of assumed salary. That assumption may well be right for your team, but it is a judgement about how much attention a cluster needs rather than something you can look up, and it deserves to be stated separately.

Two lines are easy to miss.

High availability costs more than a second server. Replication means ReplicatedMergeTree tables across multiple replicas coordinated by a ClickHouse Keeper quorum, which needs three nodes to hold a Raft majority. Every replica stores a full copy, so storage multiplies with the replica count rather than being shared.

Egress is billed by your cloud provider, not by ClickHouse. Query results leaving the region for a BI tool are charged per gigabyte, and analytics workloads move a lot of result data. Most self-hosted estimates omit it entirely, because it does not look like a database cost.

Self-hosting wins on predictable always-on load at scale, where fixed infrastructure amortises well and someone in-house already knows ClickHouse. It stops winning when the cluster is small, the team is small, or nobody owns it. This is the same tradeoff behind choosing between a managed and self-hosted database.

Managed ClickHouse pricing compared

Managed ClickHouse is priced three different ways: per compute unit-hour, per vCPU per month, and as a flat monthly plan. The pricing model matters more than the headline rate, because each one bills a different thing and the same workload lands in a different place depending on which model it sits in.

Prices verified against provider-published pricing on August 10, 2026.

The models behave differently under the same load. Consumption pricing rewards workloads that genuinely idle and penalises ones that cannot. Per-vCPU pricing is predictable but bills the same whether the cluster is busy or not. Flat plans are the easiest to budget until a usage meter tips into overage. Hourly fixed instances price like a server rather than a service, simpler to forecast and less elastic by design.

One wrinkle worth knowing: Tinybird’s plans include ingestion and backups rather than metering them separately, which changes what “cheaper” means at the same nominal price.

For a provider-by-provider breakdown rather than a pricing-model summary, our roundup of the best managed ClickHouse services goes deeper on features and support.

See the full cost comparison: Selfhost.dev vs ClickHouse Cloud.

Four ClickHouse pricing figures that are widely repeated and wrong

Four numbers circulate widely in ClickHouse pricing guides that do not match what ClickHouse publishes. Each one is checkable against their own documentation in under a minute, and each one changes a budget.

Rates verified against clickhouse.com on August 8, 2026, AWS us-east-1.

The storage error is the instructive one, because it doubles a real number by mistaking a backup for a tier premium. It runs in the same direction as the compute error: both make ClickHouse Cloud look more expensive than it is at the top end while leaving the entry price untouched, which is exactly the shape that makes a budget wrong at the point it matters.

None of this is unique to ClickHouse. Consumption pricing is only forecastable if the rate card is accurate, and rate cards move.

Where a dedicated instance changes the maths

A dedicated managed ClickHouse instance is billed by the hour for fixed hardware, so the monthly figure does not move with uptime and there is no tier to graduate into. On Selfhost.dev, a 2 vCPU, 8 GiB instance with 100 GB of storage runs $58.97 a month in Mumbai, plus about $5 for the public IPv4 added at deploy.

Here is that instance against the closest ClickHouse Cloud equivalent, priced from each provider’s own calculator.

ap-south-1 Mumbai on both sides, August 8, 2026. Backups off and egress excluded on both sides. Both configurations are a single replica with no failover.

Compute is where the difference sits: $172.37 against $46.11, roughly 3.7x. On the total it is 2.7x, because storage runs the other way. ClickHouse Cloud keeps compressed data on object storage at $27.50 per terabyte in Mumbai, while a dedicated instance uses EBS block storage, which costs more per gigabyte and does not improve with volume. At 100 GB that is $12.86 against $2.75.

The sharper comparison is against self-hosting. The same node direct from AWS is $41.80 for compute and storage. Managed, it is $58.97. The management premium is about $17 a month, against the $2,000 to $4,000 of engineering time self-hosted estimates routinely assume.

Two things about how the pricing behaves.

High availability is a choice, not a tier. Multi-AZ is an optional toggle that adds a three-node cluster with embedded ClickHouse Keeper and automatic failover. It costs extra, and it is off in the figures above.

Scaling is not a pricing event. Instance size and storage are resized on demand and billed by the hour against prepaid credits, with no tier gating and no subscription minimum. Everything pauses at a zero balance, so an idle analytics box cannot quietly run up an invoice.

The operational line is the one a managed instance removes. Provisioning, backup policies, replicas, and alert rules run from the API or from an AI editor through 161 MCP tools, which is the work self-hosted estimates price as engineering time.

Worth separating that from the query side, because a ClickHouse MCP server is read-only everywhere it exists, ClickHouse Cloud included, so it can ask questions of your data but cannot provision, resize or back up anything. There is more on managed ClickHouse on dedicated instances.

Generic pricing tells you very little about what you’ll actually pay. Calculate your costs using your instance size, storage, and region, with no account required.

Launch a dedicated ClickHouse instance in minutes.

Which ClickHouse option is cheapest for your workload

There is no single cheapest way to run ClickHouse, because the two variables that decide it pull against each other. Uptime favours fixed pricing and data volume favours object storage, so the answer depends on which one dominates your workload.

Where ClickHouse Cloud is genuinely the better buy is worth saying plainly. Per-minute metering with auto-pause on by default is the right model for anything that idles, and no fixed-instance product can match paying nothing for eighteen hours a day. Compressed object storage beats block storage at every volume, so a large archive with light querying belongs there.

Their trial is more generous at $300 across 30 days. And if you hold committed AWS spend, Altinity’s bring-your-own-cloud option keeps storage inside your own account where that commitment can absorb it, with Hetzner available as a cheaper substrate.

The honest summary: if your ClickHouse service runs around the clock, fixed hourly pricing wins on compute, which is most of the bill. If it idles, consumption pricing wins. If your data is large and your queries are few, object storage wins whoever runs it.

Getting data in is the next problem, and most teams are moving it out of Postgres.

Migration is usually the last thing holding a decision up. A Selfhost.dev engineer plans and runs it with you, free on every plan size, and you can run a dedicated instance alongside your existing setup and test the whole thing before cutting anything over. Start a dedicated ClickHouse instance.

The real cost of running ClickHouse 24/7

Running ClickHouse Cloud Basic around the clock costs $186.27 a month in us-east-1, and that service still cannot scale or fail over. That is the cost the $66.52 entry price hides, and it is the number to budget against if your analytics has to answer whenever someone asks.

• Choose ClickHouse Cloud if your workload genuinely idles, or if you hold a large archive that is queried rarely. Per-minute metering and auto-pause are hard to beat on anything intermittent.
• Choose a dedicated instance if the service runs always-on and you want a monthly figure that does not move with uptime or step onto a new price list.
• Self-host if the load is large and predictable and someone in-house already runs ClickHouse.

Consumption pricing is only forecastable if the rate card is right, which is why every figure here carries a date.

See what your ClickHouse deployment would cost, then launch it in minutes.

Frequently Asked Questions