Why Hetzner is half the price (and what you give up)
Hetzner CPX21 at EUR 5.83/month vs AWS m6i.large at $70/month. What the hyperscalers charge for that you actually get on Hetzner — and what you genuinely don't.
A Hetzner CPX21 is 3 vCPU (AMD EPYC), 4 GB RAM, 80 GB NVMe, 20 TB egress, EU-Frankfurt, for €5.83/month. That's about $6.30/month at current FX.
The closest AWS equivalent — m6i.large, 2 vCPU Xeon, 8 GB RAM, plus a 100 GB gp3 EBS volume, plus 500 GB of egress per month — runs about $90/month list, or about $63/month with a 3-year Compute Savings Plan.
Hetzner is 10x cheaper on list, 7-8x cheaper after AWS discounts. This is not a small difference. So why doesn't everyone move?
What you actually get on Hetzner
- Genuinely fast hardware. CPX-line is AMD EPYC 7003 series. CCX-line (dedicated vCPU) runs EPYC 9000 series. NVMe storage is the default, not a premium tier. Network is 10 Gbit/s shared per VM.
- 20 TB of egress per VM included in the base price, then €1/TB after that. For comparison, AWS would charge $1,800 for 20 TB of egress.
- Simple pricing. Hourly billing (capped at the monthly rate). No NAT charges. No load balancer LCUs. No region pricing variance for the EU DCs.
- Snapshots, backups, floating IPs, firewall, private network all included in the core product. Object storage and load balancers are extra-but-cheap (load balancer at €5.39/month, object storage at €5.39/TB-month).
What you give up — honestly
Managed services
This is the big one. Hetzner offers VMs, dedicated servers, object storage (S3-compatible), and a load balancer. That's it.
There is no:
- Managed Postgres / MySQL (no RDS equivalent)
- Managed Redis / Memcached
- Managed Kafka
- Managed Kubernetes (only k3s/k8s self-installed, though Hetzner Cloud has solid Terraform support)
- Managed elasticsearch / OpenSearch
- Managed data warehouse
- Serverless compute (no Lambda / Cloud Run equivalent)
- Managed identity / SSO
- Managed observability (no CloudWatch / Datadog-style native stack)
If your architecture leans heavily on managed services, the labour cost of replacing them — running your own Postgres with point-in-time recovery, your own Kubernetes control plane, your own observability — frequently eats the savings.
If your architecture is "Linux box + Docker + Postgres + Nginx", Hetzner is a no-brainer.
Region map
Hetzner Cloud has six DCs as of 2026: Falkenstein, Nuremberg, Helsinki, Ashburn (US East), Hillsboro (US West), and Singapore. That's it. No South America, no Australia, no India, no Africa, no Middle East.
If you serve latency-sensitive traffic globally, you need either an edge layer in front (Cloudflare, Fastly) or a different provider for those regions. Vultr with its 32 DCs is the usual choice for geographic breadth at low cost.
Enterprise compliance and procurement
Hetzner is ISO 27001 certified and GDPR-friendly, but it doesn't have FedRAMP, doesn't have a marketplace contract on AWS Marketplace, doesn't have an enterprise sales team that will show up to your offsite. Procurement at large enterprises often won't go near it for that reason.
Support model
Hetzner support is email-based, response within hours, no 24/7 phone for the cloud product. For most engineering teams this is fine. For a regulated workload with a 15-minute SLA on response, it's a non-starter.
Where Hetzner makes sense
- Bootstrapped SaaS running its production cluster. The cost difference between $100/month and $1000/month on infra is existential at <$10K MRR.
- Web hosting, headless CMS, e-commerce backends. Predictable load, low operational complexity.
- CI/CD runners. Hetzner is the cheapest way to run self-hosted GitHub Actions runners by a wide margin.
- Dev and staging environments even if production lives on AWS. Cuts the non-prod bill 70-80%.
- Batch workloads / data pipelines that don't need managed BigQuery / Redshift.
Where it doesn't
- You're a regulated enterprise (PCI-DSS Level 1, HIPAA, FedRAMP) that needs the certifications.
- Your architecture is "everything managed" — RDS, Lambda, DynamoDB, EventBridge. Migration cost > savings.
- You need to be in regions Hetzner doesn't serve.
- Your team genuinely doesn't have the ops capacity to run Postgres, Redis, k8s themselves.
The "60% on Hetzner, 40% on hyperscaler" pattern
The most pragmatic setup I've seen at growth-stage companies: stateless web tier + workers + caches on Hetzner; stateful managed services (RDS, ElastiCache, OpenSearch) on AWS; tied together via Tailscale or WireGuard. You get 60-70% of the savings without giving up the managed-service crutches.
For a direct catalog comparison see AWS vs Hetzner side-by-side or run your own workload through the TCO calculator. The pricing gap is real, and so are the tradeoffs.
External reading: Hetzner Cloud pricing and the AWS Cost Management blog for the other side.