AWS and Azure Are at Least 4x–10x More Expensive Than Hetzner

Original Article ↗ Hacker News Discussion ↗

Cloud engineers compare low-cost providers like Hetzner and OVH with hyperscalers such as AWS, Azure, and GCP, noting that raw compute and bandwidth on the big clouds can be 2–10x more expensive, especially once egress fees are included. Many argue that for simpler or smaller workloads, especially those that are VM-based and not mission-critical, cheaper hosts can deliver massive savings, while others emphasize that managed services, global regions, compliance, support, and reduced ops burden often justify hyperscaler pricing. A recurring theme is total cost of ownership: whether it’s cheaper in the long run to pay cloud premiums for higher-level services and reliability, or to invest engineering time in running databases, queues, and Kubernetes yourself on bare metal or budget VPS platforms.

Cost comparisons and pricing assumptions

  • Many note Hetzner (especially bare metal) is dramatically cheaper than AWS/Azure/GCP — often 4–10x on raw compute, and traffic can be ~100x cheaper than Azure.
  • Others argue realistic AWS/Azure pricing uses 1–3 year reservations/savings plans, bringing VMs to ~2x Hetzner, not 10x.
  • Some say DigitalOcean/Linode are roughly in Hetzner’s ballpark; others claim Hetzner is still significantly cheaper.
  • Several anecdotes of 70–80% cost reductions or “saved millions” after moving workloads off AWS to Hetzner/OVH.

Managed services vs “dumb VMs”

  • A major theme: hyperscalers are not just VM providers; the real value is managed databases, queues, analytics, storage, serverless, and platform services.
  • For many startups and enterprises, paying for RDS/DynamoDB/SQS/S3/etc. plus 24/7 support is cheaper than hiring ops staff and reinventing those services.
  • Counterpoint: you can self-host with Kubernetes and operators (“half‑managed”), getting ~80–90% of the functionality far cheaper, especially at scale.
  • Others argue most small apps just need “a couple of VMs and Postgres,” and complex cloud stacks are often premature.

Operational complexity and lock‑in

  • Several complain that savings plans and reservations reintroduce old-style capacity planning and vendor lock-in, undermining the original “pay for what you use” cloud pitch.
  • Some stress that once you are deep into many 3‑letter AWS services, migrating off becomes very hard and expensive.

Reliability, HA, and support

  • Hyperscalers provide automatic instance replacement, multi‑AZ/region primitives, and strong support (TAMs, engineers, cost guidance).
  • Others note you still must design for fault tolerance; bare metal can also be made highly available with proper architecture.
  • Example: a long outage at a non‑cloud provider due to hardware/network issues is cited as a tradeoff; critics respond that similar outages exist in clouds, and HA is solvable anywhere.

Traffic, locations, and use cases

  • Hetzner’s generous or unmetered bandwidth is a major advantage for high‑egress workloads; multiple commenters migrated only bandwidth-heavy components.
  • Hetzner now has EU, US, and Singapore regions but fewer locations and fewer “one-click” services.
  • Consensus: use the cheapest option that fits your real needs; for simple, low‑criticality workloads Hetzner (or similar) often wins, while complex, regulated, or fast‑moving orgs often prefer AWS/Azure/GCP.