ClickHouse Cloud vs Self-Hosted: Which Is Right for You?
A detailed cost and feature comparison between ClickHouse Cloud and self-hosted ClickHouse. Real pricing breakdowns, operational trade-offs, and a decision framework.
ClickHouse is one of the fastest analytical databases available, and the question of where to run it comes up early in every adoption. ClickHouse Cloud offers a fully managed experience. Self-hosting gives you full control over infrastructure and costs. Both are valid depending on your scale, budget, and team capacity.
This guide breaks down real costs, features, and operational trade-offs between the two approaches with concrete numbers.
ClickHouse Cloud Overview
ClickHouse Cloud is the official managed service from ClickHouse, Inc., running on AWS and GCP (Azure in select regions). It handles provisioning, replication, upgrades, backups, and scaling.
Pricing Tiers
ClickHouse Cloud uses consumption-based pricing across three tiers:
Development -- Starting around $99/month. Two replicas on shared compute. Intended for prototyping and light workloads. Performance varies with co-tenant load.
Production -- Starting around $451/month for a minimal configuration (24 GiB RAM, 6 vCPU, 3 nodes). Dedicated compute, automatic backups, and an SLA. A moderate setup (96 GiB RAM, 24 vCPU, 1 TB storage) typically lands between $1,500 and $3,000/month depending on query volume.
Dedicated -- Custom pricing for large deployments with dedicated infrastructure, VPC peering, and premium support. Expect $5,000+/month as a baseline.
All tiers charge separately for compute (per GiB of RAM, per hour), storage (~$0.024/GB/month compressed), and data transfer (egress fees vary by region). The consumption model means your bill fluctuates -- heavy analytical queries can spike compute costs well above baseline.
Self-Hosted ClickHouse Overview
Self-hosting means running ClickHouse on your own servers. A production cluster requires:
- ClickHouse nodes (minimum 3 for replication and Keeper quorum)
- ClickHouse Keeper for distributed coordination (co-located on same nodes)
- CHProxy for load balancing and connection management
- Backup tooling (clickhouse-backup to S3-compatible storage)
- Monitoring (Prometheus + Grafana)
- Firewall and network configuration
You handle provisioning, configuration, upgrades, backup validation, and incident response. The benefit is predictable costs and full control.
Detailed Cost Comparison
The following tables use 2026 pricing. ClickHouse Cloud prices are based on published rates for AWS regions. Self-hosted prices use Hetzner dedicated servers and AWS EC2 to show both ends of the cost spectrum.
Small Workload: Development and Early Analytics
A team ingesting 5-20 GB/month of event data, running dashboards during business hours.
| Component | ClickHouse Cloud (Dev) | Self-Hosted: Hetzner (3 nodes) | Self-Hosted: AWS (3 nodes) |
|---|---|---|---|
| Compute | $99/mo (shared, 2 replicas) | $30/mo (3x CPX31: 4 vCPU, 16 GB) | $315/mo (3x r6i.large) |
| Storage (100 GB) | ~$2.40/mo | Included | $24/mo (gp3) |
| Backups | Included | $3/mo | $5/mo |
| Data transfer (50 GB) | ~$4/mo | Included | $4.50/mo |
| Monthly total | ~$105 | ~$33 | ~$349 |
| Annual total | ~$1,260 | ~$396 | ~$4,188 |
At this scale, ClickHouse Cloud is competitive. The savings from Hetzner self-hosting are modest (~$860/year), and the managed service eliminates operational work. Self-hosting on AWS is actually more expensive at this tier.
Verdict: ClickHouse Cloud is reasonable unless you already run self-hosted infrastructure on affordable providers.
Medium Workload: Production Analytics
A team ingesting 50-200 GB/month with production dashboards, reports, and frequent ad-hoc queries.
| Component | ClickHouse Cloud (Prod) | Self-Hosted: Hetzner (3 nodes) | Self-Hosted: AWS (3 nodes) |
|---|---|---|---|
| Compute | $1,200-2,000/mo (96 GiB, 24 vCPU) | $135/mo (3x AX52: 12-core, 64 GB, NVMe) | $900/mo (3x r6i.2xlarge) |
| Storage (1 TB) | ~$24/mo | Included | $96/mo (gp3) |
| Backups | Included | $10/mo | $23/mo |
| Query overages | $100-500/mo | $0 | $0 |
| Data transfer (200 GB) | ~$16/mo | Included | $18/mo |
| Monthly total | ~$1,340-2,540 | ~$145 | ~$1,037 |
| Annual total | ~$16,080-30,480 | ~$1,740 | ~$12,444 |
Self-hosting on Hetzner saves $14,000-29,000/year. Bare-metal NVMe servers also deliver better I/O performance than the cloud instances underpinning ClickHouse Cloud.
Verdict: Self-hosting saves significant money. The question is whether your team can operate the cluster.
Large Workload: High-Volume Analytics Pipeline
A team ingesting 500 GB-2 TB/month with continuous analytics, real-time dashboards, and multiple consumer teams.
| Component | ClickHouse Cloud (Prod, scaled) | Self-Hosted: Hetzner (6 nodes) | Self-Hosted: AWS (6 nodes) |
|---|---|---|---|
| Compute | $4,000-8,000/mo (6+ nodes, 384+ GiB) | $390/mo (6x AX102: 16-core, 128 GB, NVMe) | $2,640/mo (6x r6i.4xlarge) |
| Storage (5 TB) | ~$120/mo | Included | $480/mo |
| Backups | Included | $30/mo | $60/mo |
| Query overages | $500-2,000/mo | $0 | $0 |
| Data transfer (500 GB) | ~$40/mo | Included | $45/mo |
| Monthly total | ~$4,660-10,160 | ~$420 | ~$3,225 |
| Annual total | ~$55,920-121,920 | ~$5,040 | ~$38,700 |
Self-hosting on Hetzner saves $50,000-117,000/year. The gap justifies hiring dedicated infrastructure capacity.
Feature Comparison
| Feature | ClickHouse Cloud | Self-Hosted |
|---|---|---|
| Setup time | Minutes | Hours (manual) / ~15 min (with tooling) |
| Automatic backups | Yes | Manual setup (clickhouse-backup + cron) |
| Point-in-time recovery | Yes (production tier) | Depends on backup frequency |
| Automatic scaling | Yes (vertical) | Manual |
| Version upgrades | Managed | Manual rolling upgrades |
| Monitoring | Built-in | Self-hosted Prometheus + Grafana |
| VPC / private networking | Dedicated tier only | Full control from day one |
| Custom configuration | Limited | Full access to all settings |
| Data residency | Limited to available regions | Any datacenter worldwide |
| Support SLA | Included | Community or paid Altinity support |
ClickHouse Cloud wins on zero operational overhead, automatic scaling for unpredictable workloads, built-in observability, and managed upgrades with no downtime.
Self-hosting wins on predictable costs, better price-performance on bare metal, full configuration control (merge tree tuning, compression codecs, disk layouts), no vendor lock-in, data sovereignty, and no egress fees on providers like Hetzner.
Hidden Costs
Hidden Costs of ClickHouse Cloud
Query compute variability. An unoptimized query from any team member can spike your bill. There is no per-query cost ceiling, making budgeting difficult across multiple teams.
Egress fees. Pulling large result sets, feeding dashboards, or running ETL pipelines that read from ClickHouse all incur transfer charges that compound over time.
Tier upgrade pressure. The development tier's shared compute leads to unpredictable latency. The jump to production pricing ($99 to $451+/month) is steep and often unavoidable as workloads grow.
Limited configuration. Some ClickHouse settings are not exposed. If your workload needs specific merge tree tuning or disk layout configurations, you may be unable to apply them.
Hidden Costs of Self-Hosting
Engineering time. A production cluster with Keeper, CHProxy, backups, and monitoring takes 1-3 days to set up manually. At $150/hour, that is $1,200-3,600 in labor before serving a single query.
Ongoing maintenance. Version upgrades, Keeper quorum debugging, and disk space management during large merges require 2-4 hours/month of engineering time.
Backup validation. Setting up backups is not the same as having working backups. You must periodically test restores.
Knowledge concentration risk. If the engineer who built your ClickHouse deployment leaves, you have a knowledge transfer problem. Thorough documentation is essential.
Security patching. OS updates, ClickHouse patches, and network security are your responsibility.
Decision Framework
Choose ClickHouse Cloud when:
- Your team is small (fewer than 5 engineers) with no infrastructure specialists.
- You are evaluating ClickHouse and want to validate before committing to operations.
- Monthly ClickHouse spend is under $500 -- savings from self-hosting are small relative to operational effort.
- You need automatic scaling for unpredictable workloads.
- Compliance certifications (SOC 2, HIPAA) are easier via a managed provider.
Choose self-hosting when:
- Managed service costs exceed $1,000/month -- the savings justify operational investment.
- You already manage self-hosted infrastructure with monitoring, alerting, and on-call.
- You need full configuration control or extensions ClickHouse Cloud does not expose.
- Data residency requirements rule out ClickHouse Cloud's available regions.
- You need predictable costs without consumption-based billing variance.
- You run on bare-metal providers where price-performance is 10-20x better.
Reducing the Operational Burden with sshploy
The strongest argument against self-hosting is setup complexity. A production cluster involves ClickHouse nodes, Keeper coordination, CHProxy load balancing, automated backups, and a monitoring stack. Done manually, this takes days and leaves room for misconfiguration.
sshploy deploys a complete ClickHouse cluster to your servers via SSH using production-tested Ansible playbooks. You select your topology (shards, replicas, Keeper placement), point it at your servers, and the full stack -- ClickHouse, Keeper, CHProxy, backups, monitoring, firewall rules -- deploys in under 15 minutes with production best practices applied automatically. This turns the "1-3 days of setup" into minutes, making the self-hosting cost comparison even more favorable. sshploy works with any SSH-accessible provider: Hetzner, OVH, Vultr, DigitalOcean, AWS, or on-premises hardware.
FAQ
How much does ClickHouse Cloud actually cost for production?
For a typical production workload (3 nodes, 96 GiB RAM, 1 TB storage, moderate queries), expect $1,500-3,000/month. The exact cost depends on query patterns since compute is billed on consumption. Teams with heavy analytical queries or many concurrent dashboards often land at the higher end. The $99/month development tier works for prototyping but not production due to shared compute.
Can I migrate from ClickHouse Cloud to self-hosted?
Yes. Use clickhouse-backup to export data, or ClickHouse's remote() and remoteSecure() table functions to copy data directly between clusters. Schema migration is straightforward since ClickHouse Cloud uses standard SQL. Plan for a migration window proportional to your data volume -- roughly 100-500 GB/hour depending on network bandwidth.
Is self-hosted ClickHouse reliable enough for production?
Yes. A properly replicated cluster (3 nodes with Keeper quorum) handles node failures automatically -- remaining replicas continue serving reads and writes while the failed node resynchronizes on recovery. Cloudflare, Uber, and eBay run self-hosted ClickHouse at massive scale. The reliability risk lies in configuration quality and operational practices, not the software.
What is the minimum team size for self-hosting?
You need at least one engineer comfortable with Linux administration and willing to learn ClickHouse operations. With tooling like sshploy handling deployment, ongoing work is modest: monitoring dashboards, validating backups, and performing version upgrades every few months. One senior engineer can manage a cluster alongside other responsibilities.
How do ClickHouse Cloud and Altinity Cloud compare?
ClickHouse Cloud uses consumption-based pricing on shared infrastructure. Altinity Cloud offers a bring-your-own-cloud model where clusters run in your AWS or GCP account, giving more control over infrastructure and costs. Altinity is generally more cost-effective for larger deployments with deeper configuration access. It is a solid intermediate option between fully managed and fully self-hosted.
Ready to deploy?
Skip the manual setup. sshploy handles the entire deployment for you.
Deploy a ClickHouse ClusterRelated guides
How to Deploy a ClickHouse Cluster: Complete Setup Guide
Learn how to deploy a production-ready ClickHouse cluster with sharding, replication, and ClickHouse Keeper. Covers architecture decisions, hardware sizing, and step-by-step configuration.
Self-Hosting vs Managed Databases: A Practical Cost & Control Comparison
Compare self-hosted and managed database services across cost, control, performance, and operational overhead. Includes real pricing breakdowns for PostgreSQL, ClickHouse, and Redis.