Cloud strategy

AWS or Hetzner? How we decide where a workload runs

We run production systems on both. Here is how we choose, and how the two work together.

AWS or Hetzner? How we decide where a workload runs

The short answer

Use AWS where you need managed services, fine-grained security controls and a region close to Zimbabwe. Use Hetzner where you need plain compute at low cost, for development, testing, internal tools or a warm standby. Many estates do best with both, joined by private VPNs.

Where AWS earns its price

  • Managed building blocks. Managed databases, queues, email, identity, AI models through Amazon Bedrock and container services such as Amazon EKS save you from running them yourself.
  • Security and audit controls. IAM, WAF, GuardDuty, CloudTrail and AWS Backup give the evidence auditors and POTRAZ inspections ask for.
  • Location. The Cape Town region (af-south-1) is the closest AWS region to Zimbabwe.

Where Hetzner makes sense

  • Cost. Hetzner servers are inexpensive, which suits environments that do not need managed services.
  • Kubernetes you run yourself. With Rancher or plain Kubernetes, Hetzner handles internal platforms, registries and staging well.
  • Standby capacity. A warm copy of production on Hetzner, kept ready, gives you somewhere to switch traffic if your primary platform fails.

We have run production workloads on Hetzner for years alongside AWS, so we are comfortable recommending either.

Running both together

  1. Private networking first. Connect Hetzner to AWS with a site-to-site VPN into a Transit Gateway, so internal tools never need public addresses.
  2. Keep internal tools internal. Registries, Rancher, Git servers, dashboards and password vaults should be reachable only over VPN.
  3. Automate certificates. When your DNS is on Cloudflare, Hetzner's managed certificates cannot validate your domain. Issue certificates with cert-manager using Cloudflare DNS validation and sync them to Hetzner load balancers automatically, or they will expire on you.
  4. Keep the standby honest. A failover site is only useful if its certificates, images and configuration are kept current.

A simple decision table

WorkloadUsually
Customer-facing production with personal dataAWS, with WAF, backups and audit logging
Development, testing and preview environmentsHetzner
Internal tools: registry, Git, dashboardsEither, behind VPN
Failover for productionHetzner, kept warm
AI workloads using foundation modelsAWS (Amazon Bedrock)

Check prices and the current service list for both providers before you commit; both change often.

This guide is general information from an engineering perspective, not legal or financial advice.

Harare city skyline

Tell us what you are running today.

Send a short note about your systems and what is getting in the way. You will hear back from the engineer who would do the work.